Double-walled carbon nanotube processing.

www.advmat.de www.MaterialsViews.com

REVIEW

Double-Walled Carbon Nanotube Processing Katherine E. Moore, Daniel D. Tune, and Benjamin S. Flavel*

Carbon nanotubes are an allotrope of carbon having the form of hollow cylinders composed of rolled-up sheets of graphene, which is itself an extended hexagonal lattice of purely

sp2-bonded carbons. Strictly speaking, carbon nanotubes are either single-walled (SWCNTs) or multi-walled (MWCNTs), but for many reasons double-walled carbon nanotubes (DWCNTs) are prominent in the literature. A single carbon cylinder, be it an SWCNT or a component wall of a DWCNT or MWCNT, can be completely described, except for its length, by an intrinsic geometric property, Ch, known as the chiral vector.[1,2] The chiral vector is defined by the equation Ch = na1 + ma2 where the integers (n,m) are the number of steps along the zig-zag carbon bonds and a1, a2 are the graphene lattice basis vectors in real space (Figure 1a). The chiral vector makes an angle, θ, known as the chiral angle, with the zig-zag or a1 direction. This angle determines the amount of “twist” in the nanotube and two limiting cases exist where the chiral angle is at 0° and 30°. These are known as zig-zag (0°) and armchair (30°) based on the geometry of the carbon bonds around the circumference of the nanotube (Figure 1b,c). All other conformations in which the C–C bonds lie at angles 0° < θ < 30° are known as chiral (Figure 1d). Because the n,m integers completely describe nanotube chirality, they also determine the electronic band structure. Thus, it is the chirality that has the most impact on the optical and electronic properties of carbon nanotubes. In particular, a slight change of the chiral angle yields nanotubes that are metallic conductors, low bandgap or high bandgap semiconductors. In the case of semiconductors, the bandgap energy is inversely dependent on the nanotube diameter. The armchair nanotubes are the only type that are intrinsically metallic (zero bandgap), although approximately a third of the zig-zag nanotubes also exhibit metallic properties at room temperature because the bandgap is smaller than the thermal energy, kBT, allowing thermal excitation of carriers into

Dr. K. E. Moore, Dr. D. D. Tune Centre for Nanoscale Science and Technology School of Chemical and Physical Sciences Flinders University Adelaide 5042, Australia

Dr. K. E. Moore, Dr. D. D. Tune, Dr. B. S. Flavel Institute of Nanotechnology Karlsruhe Institute of Technology 76021 Karlsruhe, Germany E-mail: [email protected]

Single-walled carbon nanotubes (SWCNTs) have been the focus of intense research, and the body of literature continues to grow exponentially, despite more than two decades having passed since the first reports. As well as extensive studies of the fundamental properties, this has seen SWCNTs used in a plethora of applications as far ranging as microelectronics, energy storage, solar cells, and sensors, to cancer treatment, drug delivery, and neuronal interfaces. On the other hand, the properties and applications of double-walled carbon nanotubes (DWCNTs) have remained relatively underexplored. This is despite DWCNTs not only sharing many of the same unique characteristics of their single-walled counterparts, but also possessing an additional suite of potentially advantageous properties arising due to the presence of the second wall and the often complex inter-wall interactions that arise. For example, it is envisaged that the outer wall can be selectively functionalized whilst still leaving the inner wall in its pristine state and available for signal transduction. A similar situation arises in DWCNT field effect transistors (FETs), where the outer wall can provide a convenient degree of chemical shielding of the inner wall from the external environment, allowing the excellent transconductance properties of the pristine nanotubes to be more fully exploited. Additionally, DWCNTs should also offer unique opportunities to further the fundamental understanding of the inter-wall interactions within and between carbon nanotubes. However, the realization of these goals has so far been limited by the same challenge experienced by the SWCNT field until recent years, namely, the inherent heterogeneity of raw, as-produced DWCNT material. As such, there is now an emerging field of research regarding DWCNT processing that focuses on the preparation of material of defined length, diameter and electronic type, and which is rapidly building upon the experience gained by the broader SWCNT community. This review describes the background of the field, summarizing some relevant theory and the available synthesis and purification routes; then provides a thorough synopsis of the current state-of-the-art in DWCNT sorting methodologies, outlines contemporary challenges in the field, and discusses the outlook for various potential applications of the resulting material.

1. Background

DOI: 10.1002/adma.201405686

Adv. Mater. 2015, 27, 3105–3137

© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

wileyonlinelibrary.com

3105

www.advmat.de

REVIEW

www.MaterialsViews.com

3106

the conduction band. All chiral nanotubes and the remaining two thirds of zig-zag nanotubes are therefore semiconducting at room temperature. This leads to the observation that approximately 60% of all nanotube chiralities are semiconducting with the remaining 40% metallic.[3] Carbon nanotubes are so small that they exhibit quasi-1D properties, with the confinement of electrons into allowed momentum states giving rise to van Hove singularities in their electronic density of states (DOS) (Figure 1e,f). In bulk semiconducting materials, the energy gap between the valance and conduction bands defines the absorption onset wavelength of the optical spectra, but in the case of carbon nanotubes the discrete electronic transitions in their DOS produce a series of characteristic peaks in the optical absorption spectra at correlated energies (Figure 1g). In the case of bulk semiconducting materials, the single-particle model (an electron absorbs a photon and thus moves from a lower energy level to a higher one) provides a reasonably close approximation of empirical observations. However, the 1D confinement of electrons in carbon nanotubes causes a significantly higher electron–hole binding energy (eh) and electron–electron repulsion (ee), meaning that true consideration of the optical absorption spectra of nanotubes must be entirely excitonic in nature (light energy is absorbed by the material creating excited states a.k.a. bound electron–hole pairs) with the many-body effects (eh and ee) included for an accurate description.[4] This means that the actual light energy required to produce a “vi–ci” transition is (Eci −Evi) + binding energy (eh) − self energy (ee), and leads to the experimental finding that E22/E11 tends to a value of 1.8 with decreasing tube diameter, rather than the value of 2 as would be expected from density functional theory (DFT) calculations alone.[5] The excitonic nature of the optical properties is an important consideration for specialists however, in the interests of simplicity, we present only the single-particle approximation throughout this review as this is sufficient to appreciate the concepts. There are two other important processes in nanotubes which, in addition to optical absorption spectroscopy, provide vital information regarding nanotube type and character. These are photoluminescence (PL), in which semiconducting nanotubes are excited by an S22 photon which then decays emitting a detectable S11 photon and allowing chiral identification, and Raman spectroscopy, in which the magnitude of the shift in Raman scattered light, which is exquisitely sensitive to the strong vibrational phonon modes present in the lattice of carbon nanotubes, is plotted to yield a rich variety of structural information. Detailed treatments of these phenomena are readily available in the literature and, in the case of DWCNTs, we refer the reader to Pfeiffer et al.,[6] Shen et al.[7] and Kim et al.[8] DWCNTs consist of two coaxially aligned SWCNTs and, as such, they share many of the attractive properties of SWCNTs,[9,10] but, owing to the presence of a second wall, are more physically robust and electronically more complex.[11] Similar to SWCNTs, DWCNTs are uniquely characterized by the chiral indices of the constituent inner and outer walls (ni,mi)@(no,mo), where each wall can be either semiconducting (S) or metallic (M) depending on its chiral index (Figure 1h).[12] This gives rise to four possible combinations of inner@ outer wall; namely, M@M, M@S, S@M, and S@S. This is in


Katherine E. Moore has a B.Sc. with First Class Honors and a Ph.D. in Nanotechnology from the Flinders University of South Australia. During her postgraduate degree she received many prestigious awards and grants and her work has been featured in several nanotechnology news platforms. Her research interests include the separation of carbon nanotubes, nanotube sensors, and single nanotube electronic devices. Daniel D. Tune holds a B.Sc. and Ph.D. in Nanotechnology from the Flinders University of South Australia and is the 2014 recipient of the South Australian Science Excellence Award (PhD category). Daniel is currently working as a postdoctoral researcher in the Carbon Nanotube Solar Cells and Sensors group at the Institute of Nanotechnology, Karlsruhe Institute of Technology. His interests are in renewable energy, nanoelectronics, and composite nanostructures, with a focus on the understanding of carbon nanotubes, fullerenes and graphene, and the application of this in the development of new and improved devices and materials. Benjamin S. Flavel obtained both a B.Sc. (Hons) and Ph.D. in Nanotechnology at the Flinders University of South Australia. He has since received research fellowships from the Australian Government's Endeavour Program, the Alexander von Humboldt Foundation and is currently a Junior Research Group Leader at the Institute of Nanotechnology, Karlsruhe Institute of Technology under the German Research Foundation's Emmy Noether Program. His research interests include the separation of carbon nanotubes and their integration into devices for optics, sensing, and energy.

contrast to MWCNTs, which are all metallic due to the complex inter-wall coupling. Because DWCNTs represent the simplest form of a MWCNT, they are interesting for investigation of inter-wall coupling, which has a significant influence on the electronic states of the nanotubes and is highly dependent on


Adv. Mater. 2015, 27, 3105–3137


REVIEW Figure 1. a) Unrolled graphene sheet showing the geometry of the (6,3) nanotube where the vectors OA and OB define the chiral and translational vectors Ch and T, respectively, and the rectangle OAB’B defines the nanotube's unit cell. O, A, B, and B’ are reference atoms, a1 and a2 are the graphene lattice basis vectors and θ is the chiral angle. b–d) Examples of the three classes of nanotube sidewall; zig-zag, armchair, and chiral. e,f) Schematics of the electronic density of states of semiconducting and metallic (zig-zag) nanotubes showing the valence and conduction states, vi and ci, and optically active electronic transitions, Sii and Mii. g) Optical absorption spectrum of the (6,5) semiconducting nanotube showing the S11 and S22 absorption features. h) Schematic of a double-walled carbon nanotube. i,j) Representations of commensurate and incommensurate lattice stacking.

whether the lattice stacking of the two walls is commensurate or incommensurate.[10,13–18] A DWCNT is commensurate if the ratio between the unit cell lengths of the two walls is a rational number and, as a result, the DWCNT has a periodic lattice structure (Figure 1i). In an incommensurate DWCNT, the ratio is an irrational number and the nanotube experiences broken symmetry (Figure 1j). This causes reduced inter-wall coupling because the inter-wall transfer at each lattice site oscillates around zero in a quasiperiodic manner, resulting in a net interference that is destructive.[19] In general, the extra Coulomb interaction due to inter-wall coupling reduces the electron– hole binding energy and thus lowers the optical bandgap (red shifting the absorption features) relative to the same (n,m) isolated tube; however, as discussed later, there are some wall configurations in which inter-wall coupling interactions increase the bandgap (blue shifting the absorption features). An important characteristic of DWCNTs is that their concentric structure allows for the opportunity to simultaneously exploit the chemical reactivity of the outer wall as well as the

Adv. Mater. 2015, 27, 3105–3137

excellent conductance of the unfunctionalized inner wall, making them ideal candidates for nanotube-based sensor devices. During reactions, only the outer wall is exposed to the chemical environment and can be decorated with a high density of chemical moieties.[20,21] As a result of such shielding by the outer wall, the inner wall does not suffer the drawbacks associated with functionalization, such as reduced conductivity due to degradation of the pristine sp2 hybridized framework.[22] DWCNTs exhibit great potential from both an applicationsbased and a fundamental viewpoint; however, their use has remained quite limited. This is in large part due to the lack of a method to synthesize pure, electronically well-defined raw material. While several synthesis methods, including arc discharge,[23–25] peapod growth[26,27] and catalytic chemical vapor deposition,[28] can be optimized to produce high-purity DWCNTs (up to 90%),[29] each method inevitably produces a myriad of contaminants including SWCNTs and MWCNTs, residual catalyst, amorphous carbon, and fullerenes. Purification methods exist to remove some of the unwanted



3107

www.advmat.de

REVIEW


contaminants and increase the DWCNT purity,[24,30] but they are unable to overcome the inherent dispersion of DWCNT lengths, diameters, and inner@outer wall electronic combinations. Like SWCNTs, such inhomogeneity imposes many limitations on potential applications and has spurred the development of new research focused on the preparation of DWCNT material sorted by diameter, length, and/or electronic character. This emerging field has seen the recent implementation of several strategies previously successful for SWCNT processing, such as reversible covalent modification,[31] biofunctionalization,[32] molecular nanocalipers,[33] density-gradient ultracentrifugation,[34,35] and gel permeation[36,37] with several other potential methods looking promising, such as polymer wrapping[38,39] and aqueous two-phase extraction.[40,41] As a result, well defined DWCNT material can now be used to further our fundamental knowledge of inter-wall interactions, to finally begin to realize their potential in advanced electronics and sensor applications, and to explore the rich new physics they possess. In light of this, we present a review of DWCNT processing techniques and the significance they each pose in this developing field. This review complements others in the literature that are more focused on growth mechanisms, the origin of Raman modes, the effects of temperature, pressure, and doping, the thermal and mechanical properties, photoluminescence, and outer wall selective functionalization, for which we again refer the reader to Pfeiffer et al.,[6] Shen et al.[7] and Kim et al.[8]

2. DWCNTs: Metallic or Semiconducting? Despite the ability to classify the individual electronic types of the inner and outer walls, identifying the overall electronic nature of a DWCNT is more complicated; being highly dependent on inter-wall interactions, which in turn depend on the inter-wall distance and whether the constituent nanotubes are commensurate or not.[17,42,43] Despite the fact that commensurate DWCNTs are rarely observed experimentally because it is unlikely to have two commensurate SWCNTs with the appropriate radius difference (ca. 0.33–0.41 nm[1,44] for the formation of a DWCNT,[10,13,45] they are commonly employed in theoretical studies. This is due to their defined unit cell lengths, which allow for the use of far simpler computational models and also result in very strong inter-wall coupling.[17,42,43] Indeed, theoretical calculations of commensurate DWCNTs have yielded some unexpected results, for example; a DWCNT in which both walls are semiconducting, but which exhibits overall metallic character.[17,42] Liang found that in the case of a commensurate S@S DWCNT, the energy gap scales inversely with the interwall coupling, which in turn is inversely proportional to the inter-wall distance.[14] Thus, as the inter-wall distance decreases, stronger inter-wall coupling causes the bandgap to vanish and the S@S DWCNT behaves as a metal.[14,17] Okada and Oshiyama used DFT to further investigate the curious behavior of S@S DWCNTs.[42] For a (7,0)@(16,0) DWCNT, it was calculated that the conduction band of the inner wall and valance band of the outer wall merge, forming a finite density of states at the Fermi level. However, as the outer wall diameter is increased, and thereby also the inter-wall

3108


spacing, as seen in the case of (7,0)@(17,0) and (7,0)@(19,0), the electronic structure changes to yield semi-metallic behavior. DWCNTs of increasing inner wall diameter (with respect to a constant outer wall) were also investigated using the (8,0)@ (19,0) and the (10,0)@(19,0), and both were determined to be semiconducting with finite bandgaps. Interestingly, the semiconducting (10,0)@(19,0) shares the same inter-wall spacing as the smaller (7,0)@(16,0); however, the latter is found to be metallic, highlighting the importance of the inner wall diameter and nanotube curvature on the overall electronic character of a DWCNT. Very small nanotube diameters (0.4–0.6 nm), with their greater curvature, exhibit increased σ–π rehybridization, leading to metallization.[42,46] From the work of Okada and Oshiyama,[42] it is therefore theorized that the metallization of S@S DWCNTs arises due to a combination of curvature differences between the inner and outer walls, rehybridization of the inner wall, and the inter-wall distance.[42] Likewise, Moradian et al. used DFT calculations to investigate the cases of M@S and S@M DWCNTs and found equally surprising results.[17] For a (6,0)@(20,0) M@S DWCNT, the constituent nanotubes were found to maintain their individual electronic character; however, as a whole, and in agreement with Okada and Oshiyama,[42] the DWCNT behaves as a metal due to overlap of the valence and conduction states of the component walls. Decreasing the inter-wall distance was found to cause a semiconductor-to-metal phase transition in the outer wall. Thus, both constituent walls and the overall DWCNT exhibit metallic behavior. For S@M DWCNTs, all the nanotubes investigated were determined to be metallic and, once again, the inter-wall distance was found to have an effect on the electronic character of the constituent nanotubes. For example, in (10,0)@(21,0) and (14,0)@(21,0) DWCNTs, the semiconducting inner wall becomes metallic, but in the (8,0)@ (21,0) DWCNT, where the inter-wall distance is larger, the two nanotubes exchange electronic type. Moradian et al. concluded that as the inter-wall distance is decreased, the semiconducting inner wall becomes metallic, whereas as the inter-wall distance is increased, an exchange of metallicity of the constituent nanotubes occurs.[17] Despite the increased complexity, some theoretical studies have been performed on incommensurate DWCNTs.[15,19,47,48] These show that the type of conductance is dependent upon the position of the Fermi level and on the length of the nanotubes.[47,49,50] When the Fermi energy is close to the charge neutral point, the energy levels of the constituent nanotubes are uncorrelated and coupling is weak.[49] Thus, in this energy regime, conduction is confined to one nanotube and is expected to be ballistic, as seen in the work of Ahn et al.,[49] who observed that conductance was confined to the inner wall of the (9,1)@ (17,2) DWCNT. As the Fermi energy was increased, conduction shifted from ballistic to diffusive and was distributed evenly among the inner and outer walls. While the number of experimental investigations into the transfer of charge between the constituent walls of incommensurate DWCNTs has remained small, significant insights have been gained through the work of Kalbac et al.[51] and Liu et al.[18] In the work of Kalbac et al.,[51] in situ Raman spectroelectrochemical measurements were used to observe doping behavior and the effects of charge transfer between the inner


Adv. Mater. 2015, 27, 3105–3137


Adv. Mater. 2015, 27, 3105–3137



REVIEW

and outer walls for different incommensurate M/S combinations. The results were obtained using DWCNTs with a defined outer wall electronic character, but with both metallic and semiconducting inner walls. By selecting appropriate laser excitation energies, Raman spectra of each of the four electronic combinations could be observed while the electrochemical potential (and thus Fermi level) was shifted in the positive or negative direction. When this change in Fermi level resulted in filling or depleting of an electronic state, bleaching of the Raman signal was observed.[52] As the inner wall is protected from the surrounding environment by the outer wall, it is neither in direct contact with a working electrode nor with compensating electrolyte counterions.[53] Thus, the origin of inner wall doping could only stem from transfer from its surrounding outer wall.[51] Through the use of this technique, Kalbac et al. determined that the potential required to observe charge transfer increased in the order of M@M < S@M < M@S < S@S. In each case, charge transfer was dependent upon the identity of both the outer and inner walls. For metallic outer walls, charge transfer to the inner wall could occur at small potentials close to the undoped state, Figure 2. Schematic of the electronic structures of: a) M@M, b) S@M, c) M@S and d) S@S whereas semiconducting outer walls required DWCNTs showing the Fermi level position at the point where electron transfer from the outer the first van Hove singularity to be filled. A wall to the inner wall occurs, as described in ref. [51]. Adapted with permission.[51] Copyright similar observation was made for inner walls, 2011, Wiley-VCH. with metallic nanotubes readily accepting charge from either outer wall type. Semiconducting inner walls between the inner and outer walls. Through the use of perturaccepted charge only when the filled states of the outer wall bation theory, the authors determined that any observed shift is were higher than their first van Hove singularity. This means reliant upon several factors including; the inner wall diameter that for both the S@M and S@S DWCNTs, the inner wall only and chiral angle, the outer wall diameter and chiral angle, and becomes doped when the electrochemical potential reaches that the energy level of the excitation. While these works provide of the first van Hove singularity of the inner semiconducting insight into the complex and relatively unknown interactions wall. Figure 2 shows a diagram of the respective DOS of each between the constituent walls of incommensurate DWCNTs, of the four electronic combinations of DWCNTs in the case of there is still much that is yet to be understood. shifting the electrochemical potential in the negative direction. The position of the Fermi level required for electron transfer from the outer wall to the inner wall is indicated. 3. The Characterization Problem Liu et al. also investigated incommensurate DWCNTs and have shown that under certain conditions coupling can Absorption, PL, and Raman spectroscopy are the most combe quite strong and can influence the electronic states of the monly employed characterization techniques for SWCNTs nanotube.[18] In that work, optical transitions were observed and rely on an assumed relationship between structure and properties.[56] This relationship has afforded a well-docubetween 480–750 nm (1.45–2.55 eV) for various suspended incommensurate (ni,mi)@(no,mo) combinations and the peak mented library of nanotube optical and Raman transitions that allow for the easy identification of SWCNT species.[5,56–62] positions were compared to that of the individually suspended [ 54 ] SWCNTs. Significant shifts (−50 to 200 meV) were observed As DWCNTs are a coaxial arrangement of two SWCNTs, they exhibit similar optical and Raman features to the individual across 99 different (ni,mi)@(no,mo) combinations. While a constituent nanotubes. The commonality of transitions often redshift of ca. 50 meV can be attributed to the environmental results in the application of the SWCNT library to DWCNTs; effects seen previously for SWCNTs,[55] such effects cannot however, this can be problematic. One reason is that, in the account for some of the very large and transition-dependent case of DWCNTs, convolution of the inner and outer wall variations observed, especially the significant blue shifts for optical transitions occurs, which, in combination with the prescertain transitions. Therefore, such strong variations must ence of additional SWCNTs, makes the identification of the be attributed to intrinsic effects, such as orbital hybridization

3109

www.advmat.de

REVIEW


3110

constituent wall type difficult. Unfortunately there is no clear optical absorption feature exclusively indicative of DWCNTs. This problem can be demonstrated using material that consists of DWCNTs (1.5–2 nm) as well as small- (0.8–1 nm) and large-diameter (1.5–1.8 nm) SWCNTs as impurities.[21,34,35,63] In such material, the outer wall of a DWCNT has first-, second-, and third-order semiconducting transitions (S11, S22 and S33) at ca. 2000, ca. 1150, and ca. 600 nm and inner wall transitions at ca. 1050, ca. 650, and ca. 350 nm. It is clear that the S22 of the outer wall and S11 of the inner wall overlap, and likewise the S33 of the outer wall with the S22 of the inner wall. Further complicating the absorption measurements is the presence of the metallic M11 transitions of the inner and outer walls at ca. 550 and ca. 800 nm. The optical transitions of large-diameter SWCNT contaminants also share the same region as those of the DWCNT outer walls; whilst the optical transitions of the small-diameter SWCNTs share the same region as those of the DWCNT inner walls. Figure 3 shows an example of the absorption spectra of smalldiameter SWCNTs, large-diameter SWCNTs, sorted SWCNTs, and sorted DWCNTs, where the overlap of transitions between the smalland large-diameter nanotubes can be clearly seen. Additionally, owing to strong absorption by water above 1400 nm, the S11 of the outer wall cannot be seen in a typical solution absorption measurement without the use of D2O to extend the solvent window.[64] Consequently, without the corresponding S11 transition, interpretation of S22 and S33 transitions for (n,m) identification is difficult. This issue can be resolved with the use of thin films, as seen in Figure 3. However, the spectra can differ significantly from solution measurements, where the nanotubes are present in the isolated/individualized state. This is because the excitonic properties of the nanotubes are highly sensitive to manybody interactions, Coulomb interactions, and charge transfer between adjacent nanotubes in bundles.[65] Indeed there are many examples in the literature of discrepancies between solution and substrate-bound nanotube peak positions of up to ca. 50 meV.[55] For DWCNTs, shifts in peak position are also seen on the individual nanotube level due to the effects of strong inter-wall coupling between the constituent walls.[19] Unlike SWCNTs, the optical spectra of DWCNTs are also affected by the relative


Figure 3. The absorbance spectra of films of HiPco SWCNTs (small diameter), arc-discharge (AD) SWCNTs (large diameter), sorted AD SWCNTs, and sorted DWCNTs before (solid curves) and after (dashed curves) doping with thionyl chloride. The first-, second- and third-order semiconducting optical transitions of the SWCNTs and the DWCNT outer wall (shaded red) are labelled S11, S22, and S33, and the first- and second-order metallic transitions (shaded blue) are labeled M11 and M22. For SWCNT materials, thionyl chloride doping completely suppresses the S11 transitions and significantly attenuates the S22 and M11 peaks. The sorted DWCNT film maintains several strong peaks (indicated by the arrows) in the S22, S33, and M11 regions that are attributed to inner walls of DWCNTs. Reproduced with permission.[34] Copyright 2009, Nature Publishing Group.


Adv. Mater. 2015, 27, 3105–3137


handedness of the constituent walls i.e., the S- or R- stereoisomers.[18] This arises due to obvious differences in inter-wall coupling experienced by DWCNTs of different relative handedness. Overall, the effects of inter-wall coupling can produce unusual variations in the optical spectra, making chirality assignment based on the known optical properties of SWCNTs problematic. An example of this is given in Figure 4, which shows the thirdand fourth-order optical transitions (S33 and S44) of the (15,10) nanotube, both on its own (i.e., as an SWCNT) and when contained within the outer walls of various (n,m) indices. The different inter-wall coupling within the various (15,10)@(no,mo) combinations significantly alters the peak position. Notwithstanding the variance in peak position from the expected SWCNT transitions, it is possible to perform doping experiments wherein the origin of the transitions, whether from the inner or outer wall, can be elucidated by observing the effect of the doping on the optical transitions.[34–36] In theory, only the inner wall transitions should remain active due to shielding by the outer wall, and comparison between spectra before and after doping should allow for inner and outer wall identification. However, in practice such shielding is incomplete[66] and can often introduce a new level of uncertainty. The data in Figure 3 also show the optical absorption spectra of different DWCNTs before and after chemical doping with thionyl chloride. For each nanotube of different diameter, the S11 transition is completely depleted by the doping due to a shift of the nanotube Fermi level into the highest occupied molecular orbital (HOMO) as a result of the withdrawal of electrons by the powerful oxidizer.[34–36] In the case of the large-diameter nanotubes (denoted “AD” for the arc-discharge method of synthesis), significant attenuation of the S22 transition also occurs, and in

Adv. Mater. 2015, 27, 3105–3137



REVIEW

Figure 4. The effect of inter-wall coupling on the optical transitions of DWCNTs. Optical resonances (S33 and S44) from the (15,10) nanotube show a significant energy shift. The amplitude of the energy shift depends sensitively on the specific optical transition and the outer nanotube species, and can be either positive or negative, with a magnitude as large as 150 meV. Reproduced with permission.[18] Copyright 2014, Nature Publishing Group.

the case of DWCNTs, this attenuation of S22 reveals the underlying S11 transitions corresponding to the inner walls. Whilst this cannot be used for chirality assignment, subtraction of the inner wall spectra obtained after doping from the convolution spectra of both inner and outer walls measured before doping has been used for determination of the electronic purity.[35] In a similar manner, doping/oxidation experiments can also be used in conjunction with Raman measurements.[34–36,67] In Raman spectroscopy, the effect of inter-wall coupling can be clearly seen when comparing the SWCNT radial breathing modes (RBMs), which have a well-established relationship with the nanotube diameter, to those of DWCNTs. For a specific (ni,mi) inner wall nanotube, a cluster of RBMs with almost the same resonance energy is observed,[6] spanning an 18 cm−1 frequency range.[68] This cluster arises from the same (ni,mi) nanotube residing within multiple (no,mo) outer walls, with the RBM shift dependent upon the strength of the inter-wall interaction.[69] This makes chirality assignment based on RBM frequency unreliable. Furthermore, the RBMs of DWCNT inner walls can have significantly narrower line widths than those usually observed for similar diameter SWCNTs (typically 12 cm−1), with some DWCNT RBM line widths being as small as 0.4 cm−1.[70] This is because the inner walls can be remarkably defect-free owing to the protective nature of the outer wall, and thus exhibit very long phonon lifetimes.[6] Circumventing the aforementioned difficulties in exact chirality assignment using optical absorption or standard Raman spectroscopy, a new approach has been presented by Kalbac and co-workers to discern the ratio of single- to double-walled nanotubes using Raman spectro-electrochemistry.[71] This technique exploits the diameter dependence of the high frequency, two-phonon mode (G′), which is an overtone of the disorderinduced mode (D band).[72] While both modes demonstrate a dependence on diameter, the dependence is stronger for the G′, as is its intensity arises due to different selection rules.[73] This diameter dependence results in a DWCNT-specific doublet where the higher-frequency component corresponds to largerdiameter outer walls and the lower frequency component to smaller-diameter inner walls. Upon electrochemical doping, nanotubes in direct contact with a working electrode or with compensating electrolyte counterions (SWCNTs or DWCNT outer walls) experience an upshift in frequency as well as peak broadening. In the case of SWCNTs, all nanotubes are doped and the observed peaks are upshifted. In DWCNTs however, the inner walls are protected and only the outer-wall G’ peak is upshifted. Thus, the electrochemical doping accentuates the splitting and enables distinction of SWCNTs/outer walls from inner walls.[71] By comparing ratios of pure SWCNTs and DWCNTs and employing a model relating the peak area to the nominal composition, it is therefore possible to determine the contaminating SWCNT percentage. This technique offers the advantage of reproducible and precisely controlled doping compared with traditional chemical doping strategies[71] and, furthermore, results in a clear difference between the contribution of DWCNT inner walls and SWCNTs. In terms of obtaining structural information, scanning tunneling microscopy (STM)[74] and transmission electron microscopy (TEM) are obvious choices as they can overcome the limitations of optical and Raman spectroscopy; however,

3111

www.advmat.de

REVIEW


TEM is considered the definitive tool for DWCNT measurement. TEM can achieve quantitative characterization of the real mean diameter of both the inner and outer walls, indications of the relative concentration of DWCNTs within a mixed sample containing single- and/or multi-walled contaminants, and the real diameter distribution and its standard deviation.[75] Additionally, by combining TEM with electron diffraction it is possible to determine the (n,m) type of the two constituent nanotubes in a DWCNT.[10] Unfortunately, TEM is time consuming and does not give a complete overview of the entire nanotube population under investigation nor does it provide any electronic information. The preparation of uniform samples that accurately represent the population can be difficult; surfactants are usually required to individualized nanotubes, but the presence of surfactant can introduce sample bias due to differences in surfactant wrapping of tubes and resultant interactions with the substrate during sample preparation. Therefore, unless the prepared DWCNT sample can be reliably assumed to be extremely uniform and fully representative, calculated (n,m) indices for individual nanotubes cannot be conclusively extrapolated to the entire ensemble. It is reasonable to assume that cryoTEM,[76] in which thin slices taken directly from the frozen nanotube suspension are analyzed, may circumvent this difficulty, but the technique has not yet been widely applied to DWCNTs. Thus, as a compromise, a combination of absorption techniques in conjunction with TEM is typically used to characterize DWCNT samples.

4. Synthesis Methods 4.1. Arc Discharge The arc-discharge method was first used to prepare carbon nanotubes in 1991 by Iijima whilst attempting to synthesize C60.[77] The synthesis technique involves applying a voltage and current between two closely spaced (typically 1–2 mm), highly pure graphite electrodes in an inert atmosphere. Typically, the anode is partially hollowed and filled with carbon feedstock (usually graphitic powder) and a catalyst/promoter, of which there are many combinations depending upon the desired product. Upon application of a current (steady state or pulsed), a plasma forms between the electrodes. Random collisions between the carbon atoms, catalyst particles, and gas then result in the formation of carbonaceous material as a macroscopic deposit on the cathode, as well as the reactor walls. Fullerenes and carbon nanotubes of varying types can be found within this carbonaceous soot.[78] The type of carbon nanotube produced can be tailored to be single-, double- or multi-walled through careful control of catalyst composition, atmosphere, current/ voltage conditions, and carbon feedstock. This field is vast, but some significant reports on the optimization of these growth conditions to yield DWCNTs are now discussed. Hutchison et al.[23] were the first to selectively synthesize DWCNTs with the steady-state arc-discharge method in 2001. They used a mixture of Ni, Co, and Fe catalyst within the graphite anode in an Ar/H2 atmosphere with varying amounts of S included as a growth promoter. The resultant DWCNTs had a purity varying from ca. 10% to 70% and at the end of

3112


their study the authors concluded that the presence of S in the catalyst was not critical to the formation of DWCNTs. However, in later work by Saito et al., the presence of S as a growth promoter was concluded to be indispensable for the production of DWCNTs.[29] They also concluded that the catalyst must contain Fe and that H2 was of vital importance for DWCNT production. Upon optimizing the catalyst (FeS:NiS:CoS = 1:1:1) and the atmosphere composition, the highest purity achieved was ca. 90% DWCNTs with the average outer diameter ranging from 2 to 5 nm. The presence of halides can also improve the yield of DWCNTs, as demonstrated by Qiu et al. using steady-state arc discharge in a hydrogen atmosphere.[25] Through the introduction of chloride (specifically KCl) to the catalyst mix, they achieved an increased yield of DWCNTs (10 wt% without KCl to 50 wt% with KCl) with a purity of 90%. While an atmosphere containing H2 is now commonly employed, it is possible to achieve DWCNT growth without the presence of reactive gases. Huang et al.[79] prepared DWCNTs without H2 by customizing the shape of the cathode to a bowllike structure. Using the now well-established catalysts of Ni, Co, Fe, and S, they demonstrated the growth of DWCNTs with a purity of ca. 80%. The DWCNTs possessed improved oxidation resistance compared to conventionally prepared arc-dischargeor catalytic chemical-vapor-deposition-produced DWCNTs, owing to the large hot region within the bowl-like cathode. This hot region allowed for in situ annealing or “defect-healing” of the DWCNTs. A report of DWCNT synthesis by Sugai et al. demonstrated the first use of high-temperature pulsed arc discharge without H2.[24] Contrary to the steady-state methods, pulsed arc discharge also allows for DWCNT production without the presence of Fe and S. In this case, an Y/Ni catalyst commonly used in the formation of SWCNTs was used,[30,80] however a change from preferential formation of SWCNTs to DWCNTs occurred due to the increased temperature. It has been shown that the SWCNT diameter increases with increased temperature until 1200 ºC, at which point the nanotube reaches a critical diameter and DWCNTs become the favored product. Owing to the highly controlled arc conditions (600 µs pulse at 50 Hz), small-diameter DWCNTs (1.6–2.0 nm) could be produced. Similarly, Zhao et al. also demonstrated the growth of DWCNTs without the presence of H2 (or any expensive highpurity gases) with a steady-state arc-discharge approach.[81] In that work, Fe was used to catalyze the growth reaction with S present as a promoter, while dry air flowed throughout the reaction at reduced pressure. The authors showed that DWCNTs were the dominant product with diameters ranging between 3 and 7 nm, but that SWCNTs and triple-walled carbon nanotubes (TWCNTs) were also present, as well as a considerable amount of other carbonaceous material. Whilst the gross DWCNT purity is significantly reduced without the presence of H2 (only 20% was reported by Sugai et al.[24] and no purity was estimated by Zhao et al.,[81] such a route offers the advantages of reduced cost and a simpler growth process, and minimizes the dangers associated with large-scale application of H2 gas. In the future, the issue of low purity may be addressed through further optimization or post-growth purification strategies. Although catalyst composition and reaction atmosphere are critically important for DWCNT growth and much work to date has been focused on these aspects, the literature also


Adv. Mater. 2015, 27, 3105–3137


4.2. Peapod Growth DWCNTs can be grown by the so-called “peapod” method by first encapsulating a precursor material within an SWCNT and subsequently treating it to induce coalescence and thereby form an inner wall. The first report of this technique was by Smith et al.,[86] who observed that during the pulsed vaporization of graphite, a technique previously reported to synthesize SWCNTs and fullerenes,[87] C60 and C70 can become trapped inside appropriately sized SWCNTs. In situ TEM revealed that the fullerenes were deposited on the surfaces of the nanotubes from the gas phase and were seen to enter the nanotubes via the uncapped ends or through defect sites. Once inside the nanotubes, the fullerenes self-assembled into chains, called “bucky-peapods” with uniform center-to-center distances. After extended exposure to a 100 kV electron beam or, as later discovered, temperatures above 1100 ºC,[26] the fullerenes coalesced to form the inner wall of a DWCNT with a nearly uniform inter-wall spacing of 0.3 nm. An in-depth investigation into the temperature dependence of the DWCNT formation conducted by Bandow et al. determined that the process of coalescence

Adv. Mater. 2015, 27, 3105–3137

only occurs at temperatures above ca. 800 ºC, becomes better with increasing temperature, and is complete at ca. 1200 ºC.[88] The authors further report that other fullerenes, such as C76, C78, and C80, can be used as a precursor material and that the diameter of the resultant inner tube is determined solely by the diameter of the parent wall, irrespective of the fullerene size. Thus, other carbon-containing aromatic precursor materials could very well be used for DWCNT formation, provided that they can be encapsulated by the parent nanotube. Indeed, there are several reports of SWCNT encapsulation of ferrocene[89] which further enables the introduction of a metal catalyst.[90] Recently, peapod growth has been demonstrated by photoninduced decomposition of fullerenes.[27,91] With the use of a UV laser, Berd et al.[92] irradiated C60 peapods with photons of energy higher than that of the fullerene bandgap (1.7 eV for C60) resulting in C60 fragmentation.[93] Berd et al.[92] determined the optimum laser excitation was 3.7 eV, which effectively achieved coalescence of the encapsulated C60 molecules while maintaining the structural integrity of the parent nanotube. This technique was claimed to be advantageous for nanoelectronic applications as it allows for in situ growth of DWCNTs at room temperature and in ambient conditions,[27,91] but, of course, most conceivable nanoelectronics applications will be exquisitely sensitive to the type of inner wall that is grown. This means that this technique has questionable real-world applicability unless the type of nanotube that is grown can be finely controlled. The peapod growth method affords very high quality DWCNTs[70] that are clean of residual catalyst and enable the production of very thin DWCNTs (theory predicts diameters as small as 1.174 nm can be filled with C60.[94] Purities of ca. 90% can be achieved, with the only source of contaminant being SWCNTs.[92] However, the method still requires small-diameter SWCNTs to first be synthesized, purified, and carefully processed to allow for the introduction of the precursor material.[95] There is also the issue of unfilled SWCNT contamination, which must currently be addressed post synthesis (although process refinements could likely address this issue in full). Such costly, time consuming, and inherently wasteful steps add significantly to manufacturing complexity, reducing the viability of commercial application and meaning that peapod growth will likely remain limited to small-scale research applications for the foreseeable future. Furthermore, poor filling of the parent outer wall with the coalesced inner wall remains an issue in that theoretical calculations predict that when the fully packed C60 molecules inside the parent nanotube have completely transformed, only ca. 2/3 of the parent nanotube has been filled, leaving empty space.[88] However, it is conceivable that this could be a distinct advantage in certain applications requiring non-standard on-tube changes in electronic properties, which could in theory be moved back and forth along the outer wall through control of the position of the inner wall, or similarly in nano-electromechanical systems (NEMS).

REVIEW

contains studies into the effect of the composition of the carbon feedstock. Li et al. demonstrated that MWCNTs/CNF (carbon nanofibers) can be successfully used to grow high-purity DWCNTs by steady-state arc discharge in a H2 atmosphere.[82] The anodic carbon feedstock (usually high-purity graphite >99%) was substituted with the MWCNT/CNF mixture, which had a diameter range of 40–220 nm. The authors found that DWCNTs produced from MWCNTs/CNFs were of higher quality than those prepared by graphite powder using the same process and they determined the DWCNT purity to be 83% with the outer diameter ranging from 1.75 to 4.87 nm. Recently, Xu et al. demonstrated that asphalt[83] and petroleum coke[84] could also be used as the carbon feedstock in lieu of high-quality (and very expensive) graphitic powders. In both of those cases, Fe was used as the catalyst and, in the case of petroleum coke, an Ar atmosphere was shown to lead exclusively to the production of DWCNTs with diameters of 3–4.4 nm. While arc-discharge methods can produce high-purity DWCNTs (up to 90%[29]) with few structural defects, it is limited to large diameters and provides low yield. For instance, arc-discharge methods generally produce nanotubes with diameters >2 nm. Although these may be desirable for some applications, large diameters are not easily characterized using Raman or absorption spectroscopy owing to limitations in the spectral windows, and very large diameters (>4 nm) can easily collapse.[85] Currently, the mean diameter distribution can be reduced through the use of pulsed arc-discharge methods to give nanotubes with diameters of 1.6–2 nm; however, this comes at great expense to the yield, which is significantly reduced to only 10%.[30] This is quite low compared to the yields of steady-state methods employing H2, Fe, S, and halides, which produce large-diameter DWCNTs of up to 50%.[25] While the arc-discharge method may be limited to large diameters currently, future investigation into the thermodynamics and kinetics of growth, as well as the role of growth promoters, may lead to a higher level of control over the diameter.

4.3. Catalytic Chemical Vapor Deposition In catalytic chemical vapor deposition (CCVD) growth, a volatile gaseous carbon source (typically CH4, CO, or C2H2) is



3113

www.advmat.de

REVIEW


3114

decomposed at high temperature over metallic nanoparticles. The decomposed carbon atoms diffuse into the metal nanoparticles and, upon saturation, precipitate at the surface and initiate nanotube growth directly from the nanoparticle. Thus, the catalyst particles simultaneously act as nucleation points for nanotube growth and provide catalytic enhancement. The earliest report of CCVD growth was in 1976 by Oberlin et al., who decomposed benzene in a H2 atmosphere at 1100 ºC to produce nanotubes with diameters ranging from 2–50 nm, although these were not recognized as such at the time.[96] Later, Dai et al. used catalytic decomposition of CO over nanometer-sized Mo particles at 1200 ºC to form individual SWCNTs with diameters ranging between 1 and 5 nm and realized that the diameter of the catalytic particles closely correlated to the diameter of the resultant nanotubes.[97] This correlation was further investigated by Cheung et al., who synthesized nanotubes using 3, 9, and 13 nm catalyst particles with C2H2 or CH4 as the carbon source at a temperature of 800–1000 ºC.[98] It was found that when the catalyst particles were small, the resultant nanotubes were primarily SWCNTs with ca. 30% DWCNTs. By increasing the size to 9 nm, SWCNTs, DWCNTs, and MWCNTs were produced. Finally when the catalyst particles were large (13 nm), only MWCNTs were produced. The shape of the catalyst is also important, as shown by Liu et al., who grew DWCNTs from CH3 decomposition over porous Fe/MgO.[99] Through application of external pressure, the pore size was varied and the growth of the DWCNTs was found to be highly dependent on the pore size of the catalyst, with pores less than 30–50 nm producing only MWCNTs. It was proposed that the compression and small pore size resulted in deformation and agglomeration of the metal nanoparticles, yielding MWCNTs and carbon capsules. Furthermore, the pore size can physically hinder the growth of DWCNTs when the length of the DWCNTs reaches the depth of the pore. At this point growth can be terminated, extend into the catalyst structure, or buckle and change direction. Thus, large pore size or a loose stacked structure is best when using porous catalyst. However, the deliberate growth of high-purity DWCNTs is not as straightforward as simply controlling the catalyst size or shape and other factors such as catalyst composition,[100,101] temperature,[102,103] atmosphere,[104] and growth templates[105,106] also play pivotal roles. The importance of catalyst composition was investigated by Hafner et al., who synthesized SWCNTs and DWCNTs from the catalytic decomposition of both CO and C2H4 using nanometer-sized alumina and Mo as the catalyst.[102] Only SWCNTs of monodisperse diameter were produced; however, when Fe was introduced into the catalyst mixture, the reaction also produced DWCNTs, highlighting the effectiveness of Fe for catalyzing DWCNT growth. Currently, Fe remains the most commonly used catalyst owing to its catalytic activity for the decomposition and formation of metastable carbides and because carbon is able to rapidly diffuse through and over the metal surface.[107] However, there are examples of other transition metal catalysts producing DWCNTs, such as by Flahaut et al. who used MgCoO catalyst.[108] They demonstrated gram-scale growth of DWCNTs (up to 1.3 g from 10 g of catalyst) using CH4 as the feedstock by the inclusion of Mo in the catalyst mixture, which had already been shown to increase the yield of nanotube production.[109] It was later demonstrated that the preparation route


for the catalyst was of equal importance as the catalyst composition itself.[101] By altering the synthesis route of MgCoMoO from urea-based combustion to citric-acid-based combustion (a milder combustion process) the catalyst had a higher specific surface area and improved homogeneity, resulting in almost 80% DWCNT growth. Similar to the arc-discharge method, the addition of S as a growth promoter can also change the reaction preference from SWCNTs to DWCNTs. This was investigated by Ci et al. for CCVD growth from C2H2 decomposition at 900– 1100 ºC.[110] It was reported that the growth of DWCNTs was strongly dependent on S addition and, without its inclusion in the catalyst material (ferrocene), only SWCNTs were produced. This was also the first report of CCVD synthesis of DWCNTs using a “floating catalyst”, a popular variation of CCVD used for SWCNT growth, which involves subliming the catalyst into the gas phase rather than using substrate-bound form and offers the advantage of producing very long nanotube ropes.[111] Further improvements in DWCNT purity can be achieved through temperature control.[102] It is found that by simply increasing the growth temperature from 700 ºC to 850 ºC, the production of DWCNTs increased from a composition of only 30% to 70%. CCVD can be used as a means to grow DWCNTs in vertically aligned forests, which allows for their direct integration into applications such as sensors[112] and field emitters.[113–116] The first demonstration of vertically aligned DWCNTs grown on flat substrates was by Yamada et al. using water-assisted CCVD.[117] This was achieved using Fe-Al2O3 (30 nm)/SiO2 (600nm)/Si, with C2H4 as the carbon feedstock, and a DWCNT purity of 85% was produced. The high purity was attributed to enacting precise control over the Fe catalyst film thickness, which largely determines the size of the catalyst particles formed upon heating. This was further investigated by Ci et al. who produced high-purity DWCNTs (88%) using Fe films on an Al support layer with C2H4 feedstock at 700–850 ºC.[118] It was determined that an Fe film thickness of 1.5 nm deposited onto 10 nm of Al provided the greatest selectivity towards DWCNT growth. The optimized growth conditions resulted in ultra-low-density arrays due to the very large diameter of the resultant DWCNTs (7.9 nm) and the low catalyst particle density. Ultra-long, superaligned DWCNT forests have been reported by Kim et al. with lengths of 9 mm in 10 h of growth time.[119] This was achieved by first reducing the catalyst (Fe film (1 nm)/Al2O3 (30 nm)) by exposure to a He/H2 atmosphere at 750 ºC. The reduction time had a significant effect on the density and size of the catalyst particles and ultimately changed the quality and alignment of the nanotubes forests, where, at 5 min of reduction time, the catalyst formed the smallest grain size, leading to the highestquality growth. Due to CCVD’s compatibility with a wide variety of substrates, researchers are also focused on preparing aligned DWCNT arrays on substrates more easily integrated into electronics, such as silicon or gold. For instance, Chen et al. produced vertically aligned DWCNT arrays using point-arc microwave plasma CCVD to allow for direct measurement of field-emission properties.[114] In that work, a three-layered catalyst of Al2O3/Fe/Al2O3 (0.5 nm/0.8–1 nm/5 nm) was deposited onto both Cr (100 nm)/Si and SiO2 (200 nm)/Si. The trilayer catalyst provides two functions: the thicker, bottom Al2O3 layer acts as a barrier, preventing the Fe catalyst from reacting with


Adv. Mater. 2015, 27, 3105–3137


Adv. Mater. 2015, 27, 3105–3137

not exist because the probability of growing either SWCNTs or MWCNTs is never zero when the parameters are optimized to maximize growth of the other; therefore, the typical product is always a mixture of nanotube types.

REVIEW

the underlying substrate; the thin, top Al2O3 layer serves to increase the surface diffusion barrier of the Fe atoms so as to suppress aggregation. Growth of the DWCNTs occurs via the decomposition of CH4 in a hydrogen atmosphere at 600 ºC, yielding outer diameters of 2.5–3.8 nm. It was determined that the Cr/Si substrate achieved a higher nanotube density, but had a lower growth rate than the SiO2/Si substrate. This was attributed to the SiO2 accumulating heat during the plasma radiation exposure. Importantly, when growing DWCNTs for direct integration into electronic applications, good contact between the DWCNTs and underlying substrate is essential. Chen et al. determined that the presence of Cr provided less resistance (compared to the SiO2) and the DWCNT/Cr/Si surface demonstrated superior field emission. Liu et al. later showed that SiOx (30 nm) deposited onto SiO2 (100 nm)/Si surfaces could be directly used to catalyze DWCNT growth.[120] It was determined that the critical factors for growth were the thickness of the deposited SiO2 layer and pre-growth heat treatment, which both affect the size of the resultant SiOx catalyst particles and, in turn, the number of walls. Nanoparticles in the range of 3–5 nm (annealed for 10 min at 850 ºC in Ar atmosphere) were found to be the most effective for DWCNT growth, achieving a purity of 70%. Growth directly on conducting metal foils has been reported by Iijima and co-workers, with Ni-based alloys containing Cr or Fe found to be best for DWCNT and SWCNT growth.[106] In that work, an Al2O3 (30 nm)/Fe (1 nm) catalyst layer was deposited onto various metal foils and then exposed to C2H4 in a H2O/He atmosphere at 750 °C. By selectively tuning the catalyst thickness to 1.8 nm, DWCNTs were produced and measurement of the field-emission properties confirmed that good electrical contact had been established. As well, Fu et al. demonstrated that DWCNTs could be grown on catalyst (Fe/Al2O3)-coated Au films for use as field-effect emitters.[121] This was done by first reducing the catalyst in a H2 atmosphere before introducing C2H4 at 700 ºC. CCVD is the most commonly used growth method to produce DWCNTs as it offers high yields, is cost effective and controllable, and is appropriate for industrial-scale production. Furthermore, the ability to grow nanotubes on a substrate enables direct integration into some applications, as well as ease of collection. However, despite considerable advances in favoring the growth of DWCNTs over other species, CCVD growth cannot yet be tuned to generate a specific DWCNT type and instead produces a heterogeneous mixture of DWCNTs with small- and large-diameter SWCNTs and even triple- and multiwalled contaminants, as well as amorphous carbon and remnants of metal catalyst. The slow development is in part due to the time-consuming nature of the experiments, where each CCVD run can take an entire day to complete and characterize, as well as the large parameter space of the CCVD process itself. Recently, Nikoleav et al. reported the use of automated experimentation as a means to overcome this cumbersome approach and conduct over 100 water-assisted CCVD growth cycles in a single day.[122] The adaptive and rapid experimentation system, equipped with in situ Raman spectroscopy, is capable of mapping regions of selectivity toward SWCNT and MWCNT nucleation and growth in a four-dimensional parameter space from only a limited number of input experiments. From this, the authors conclude that “perfect” selectivity does

5. Processing and Sorting The field of DWCNT processing is still in relative infancy compared to the more extensively investigated field of SWCNTs. In some ways, the sorting of DWCNTs could be viewed as an extension of already-established SWCNT sorting techniques; however, even at this early stage of development, important differences are apparent and these will only become more prominent as researchers refine the specificity of their methodologies toward the ultimate goal of sorting by inner wall character. The discussion that follows is thus necessarily dominated in some parts by the descriptions of progress in SWCNT sorting. Where SWCNT sorting techniques have been utilized for DWCNTs, some notable results are presented, and in all cases the applicability and potential benefits in relation to DWCNT sorting are analyzed. Unless new techniques are developed that provide for targeted growth of highly pure DWCNTs of desired type, there is a clear need to improve the purity of as-produced DWCNT material in order to unlock their potential in useful applications. This can be achieved through several different approaches, where each approach targets certain contaminants. These approaches can be broadly grouped into two categories: purification, which involves the crude removal of amorphous carbon, fullerenes, SWCNTs, MWCNTs, and any remaining catalyst through chemical or thermal treatment; and sorting, where molecular control is enacted to finely refine the desired product according to requirements e.g., the preparation of material with exclusively semiconducting outer walls.

5.1. Purification Several treatments can be applied post-synthesis to improve the purity of raw DWCNT material by selective removal of contaminants. For example, the pulsed arc-discharge method of Sugai et al. reported only 20% DWCNT purity due to the presence of fullerenes, SWCNTs, and amorphous carbon contamination.[24] Following purification, they were able to dramatically increase the DWCNT purity to ca. 90%. In their purification strategy, the as-prepared material is first washed with CS2, a solvent known to selectively solubilize fullerenes, allowing their removal by standard laboratory separation techniques.[123] Residual catalyst particles are then dissolved via sonication in concentrated acid; in this case HCl. Treatment with various acids has been used extensively to purify SWCNTs and has been shown to effectively remove metal particles from as-grown nanotubes.[124] However, the process inevitably results in functionalization of the nanotubes to some degree, depending on the temperature and length of exposure. The amorphous carbon and SWCNT material were then removed via subsequent high temperature air oxidation at 500 °C for 1 h. As the SWCNTs have the same average diameter as the DWCNTs (ca. 2 nm), the authors conclude that the greater thermal stability of the DWCNTs originates from a



3115

www.advmat.de

REVIEW


higher degree of graphitization (with fewer defects) and from interactions between the inner and outer walls. The same group later quantified this effect by determining that under optimized conditions the oxidation rate of DWCNTs is half that of SWCNTs of the same diameter (1.6 nm); hence, SWCNTs are preferentially oxidized and removed.[30] Purities can be greatly improved (from 10% up to 90%) however some hollow and metal-filled carbon capsules remain even after HCl treatment and cannot be removed. Furthermore, it has been reported that hot air oxidation exhibits a selectivity toward metallic SWCNTs, with semiconducting species still remaining after 4 h at 420 ºC, highlighting the need for the correct oxidation temperature in DWCNT purification.[125] However, no reports of air oxidation of DWCNTs specifically measure the electronic properties of the remaining SWCNTs; thus, it is unclear how the preference for metallic oxidation affects the remaining SWCNT population. While air oxidation can effectively remove amorphous carbon and SWCNTs in optimized conditions, and offers the advantage of oxidation without the introduction of sidewall defects, it can result in destruction of the nanotubes at temperatures above 750 ºC.[126] Alternatively, refluxing in H2O2 for 12 h can also be used to oxidize unwanted SWCNTs, as demonstrated by Yoshida et al., who saw a dramatic improvement in DWCNT purity from only 10% initially to 95% after refluxing.[30] As with acid treatment, some oxygen-containing moieties are introduced at defect sites within the lattice,[127] but to a much lesser extent, and Raman studies show a dramatic decrease in the amount of amorphous carbon present.[128] Similar to hot air oxidation, the H2O2 also exhibits a preference to attack nanotubes of specific electronic character; however, in this case it is toward semiconducting character. This is evidenced by the H2O2 treatment of SWCNTs resulting in a population of ca. 80% metallic composition after ca. 1 h.[125] In the DWCNT case, refluxing in H2O2 is for a significantly increased duration and it is unclear whether or not metallic SWCNTs remain after such an extended time. However, one can conclude that, since air oxidation and H2O2 exhibit a preference toward metallic and semiconducting SWCNTs, respectively, a combination of the two may prove to be an efficient purification strategy.

5.2. Suspension All sorting strategies share one commonality, namely the requirement for the production of suspensions of individualized carbon nanotubes in either water[58,61,129,130] or organic solvents.[131,132] This is achieved through either covalent[20,31,133] or non-covalent methods.[76,129,134–136] Covalent chemistry routes involve the introduction of functional groups to the nanotube ends and sidewalls, rendering them soluble.[31,137] Such processes are extremely good at producing well-dispersed, individualized nanotube suspensions, and often also exhibit some selectivity toward certain diameters[31,138] or electronic types,[139] which can provide useful routes toward separation based on these characteristics. However, covalent functionalization is often disadvantageous as it causes disruption of the conjugated π system of the nanotube by introducing sp3 hybridization into the pristine sp2 network.[22] Non-covalent approaches include the use of surfactants such as sodium cholate (SC) or sodium

3116


dodecyl sulphate (SDS),[60–62,130,140–143] or dispersing agents such as DNA[58,144–146] and organic polymers.[131,132,147,148] Dispersion of the nanotubes in these stabilizing agents is achieved via ultrasonication and often followed by centrifugation to remove remaining bundles and residual catalyst particles.[136] For surfactant-stabilized dispersions, the surface concentration and orientation of surfactant molecules on the nanotube sidewalls is highly dependent upon the type of surfactant, the concentration in solution, and the diameter and electronic type of the nanotube.[60,62,140,149–153] Generally, the surfactant first forms a random layer and, as more surfactant molecules bind to the surface, begins to form hemimicelles. Further addition of surfactant can cause the formation of a highly packed cylindrical micelle.[154] However, this varies for different diameters, with smaller nanotube diameters exhibiting lessordered surfactant structures.[149] For example, SDS wrapping of small-diameter nanotubes (1 nm) is also disordered at low concentration but forms hemimicelles at high SDS concentration.[155,156] Experiments suggest that the correlation between nanotube structure and wrapping is due to differences in the surface π-electron states of the various SWCNT curvatures, which affect the SDS/ nanotube interaction.[60] When an SDS molecule wraps around a small-diameter nanotube with a large bond curvature, it encounters a larger energetic barrier due to bending.[156] Therefore, it is energetically more favorable for the SDS to adsorb to larger-diameter nanotubes with smaller curvatures, which is an underlying principle of several of the sorting methodologies described later.[140,155,157] The extent of SDS encapsulation is also dependent upon the electronic character, with metallic nanotubes having a higher degree of SDS wrapping than semiconducting nanotubes, owing to their higher polarizability.[61,140,152,158] Co-surfactant wrapping, which relies upon competitive non-specific binding between different surfactants e.g., SC and SDS, provides a further parameter that can be exploited for nanotube separation according to electronic character.[58,130,142,143] Suspension of nanotubes with DNA[159–161] and proteins[32,162–165] relies primarily on strong π-stacking onto the nanotube sidewalls. This effectively individualizes the nanotubes and at the same time provides a negative surface charge density due to close proximity of the phosphate backbone to the nanotube.[159] As such, DNA is particularly suitable for solubilization of DWCNTs, with several examples seen in the literature.[135,160,161,166] Using DNA, SWCNTs have been sorted by electronic type[144,159] and (n,m) species[145] through the use of ion-exchange gels. Because they contain biological elements, such suspensions are highly dependent on the pH of the solution, which must be optimized to achieve well-dispersed nanotube suspensions of high concentration. Suspension in organic solvents is used in polymer separation and will be discussed in Section 5.9. Once a good suspension of purified nanotubes has been obtained, there are several techniques that have been developed to sort them using a number of different strategies with varying degrees of complexity, specificity, and success.


Adv. Mater. 2015, 27, 3105–3137


MWCNTs were enriched in the insoluble solid. Deng et al. also demonstrated that the reduction occurs by defect activation and propagates exclusively from sp3-hybridized sites, giving rise to clusters of functional groups, and thus explaining why the amorphous carbon, which primarily consists of sp3-hybridized carbon atoms, can be eliminated so easily.[133] By further performing a subsequent 3-cycle alkylcarboxylation on the insoluble solid, the DWCNTs and MWCNTs were then selectively functionalized and water solubilized, leaving the remaining MWCNTs in the insoluble solid. A schematic illustration of the two-step diameter-dependent reaction can be seen in Figure 5. After the successful enrichment of DWCNTs, the alkyl functional groups were then removed by thermal annealing of the filtered and washed product in H2/Ar. High-resolution TEM revealed that while the final sorted product did still contain some MWCNTs (likely resulting from their very high initial composition), the carbonaceous material and SWCNTs had been successfully removed, leading to DWCNT enrichment in the sample. This work not only demonstrates the applicability of covalent chemistry to enrich DWCNTs, but does so using a straightforward and highly scalable technique, which further allows the nanotubes to regain their pristine structure. Although no enrichment by electronic character was observed, the reduction potential of nanotubes does exhibit dependence on electronic character, and even on chirality. In theory it should therefore be possible to use a stoichiometric deficiency of Na to selectively functionalize only those nanotubes with the smallest reduction potential and thus facilitate some degree of sorting.

Because various covalent modification schemes are selective to nanotube diameter, such inherently scalable approaches are highly attractive for the purification and separation of DWCNTs. One such approach demonstrated by Deng et al.[31] involves the Billups–Birch reductive alkylcarboxylation reaction. That work involved covalently functionalizing CNTs with alkylcarboxylic acids groups progressively from smaller-diameter nanotubes towards larger diameters. Addition of these groups rendered the nanotubes water soluble and the diameter-dependent nature of the reaction allowed for standard phase separation in water–hexane. This resulted in partitioning of functionalized samples into different aqueous extracts of decreasing functionalization and solubility, commensurate with increasing diameter. As well as providing such diameter-selective separation, a further advantage of this kind of approach is the reversible nature of the reactions. In the case of alkylcarboxylation, the sorted nanotubes can be returned to their pristine state through simple annealing. Because wall number is closely correlated with diameter, reductive alkylcarboxylation can also be used to selectively remove single- and multi-walled nanotubes, as well as carbonaceous by-products, from as-synthesized DWCNT material.[167] In the reductive-alkylcarboxylation method, raw DWCNT material is added to liquid NH3 containing Na to individualize the nanotubes and enable homogeneity of the subsequent reaction. When Na is added to NH3, it dissolves completely, with nitrogen lone pairs coordinating to the metal, yielding an electronic liquid comprising [Na(NH3)x]+ complexes and free electrons. When nanotubes are added to the solution, the solvated 5.4. Biofunctionalization electrons are transferred to the various nanotubes and (aromatic) carbonaceous material, providing a negative surface An example of the use of biological elements to suspend and charge and producing “nanotubide” radical anions. In this sort DWCNTs is seen in the 2011 work of Nie et al. in which way, the van der Waals forces holding the nanotubes in bunDWCNTs were non-covalently biofunctionalized with the prodles are gently overcome by strong Coulombic repulsion, thus tein, lysozyme.[32] Lysozyme is particularly desirable for this individualizing them.[31] Upon addition of 6-bromohexanoic acid, as in the work of Deng et al.,[31] (or any molecule with a purpose as it primarily consists of amine groups, which provide the nanotube with adequate water solubility,[165] and has halogen-terminated alkyl chain) a reduction reaction occurs adding alkylcarboxylic acid groups to the nanotube sidewalls. hydrophobic residues within its core that readily interact with The diameter selectivity arises due to the difference in the Fermi levels of the nanotubes, wherein smaller-diameter nanotubes (with higher Fermi levels)[168] exhibit a greater reduction potential than larger ones. Hence, the reduction by solvated electrons occurs preferentially on the smaller-diameter nanotubes and, with careful control of the reaction stoichiometry, this can then lead to preferential addition of alkylcarboxylic acid groups to the smaller nanotubes. In the work of Deng et al.,[31] three reaction cycles were completed, each starting with further addition of Na and alkylcarb- Figure 5. Schematic illustration of the two-stage enrichment of DWCNTs using diameter- and oxylate source, before phase separation in a defect-selective alkylcarboxylation. By exploiting the diameter dependence of the reductive alkylcarboxylation of carbon nanotubes, the degree of functionalization and water solubility of the water/hexane mixture. The more easily funcSWCNTs were selectively enhanced, allowing for their removal in aqueous extracts by water– tionalized SWCNTs and amorphous carbon hexane-phase extraction. Additional reaction cycles were performed on the insoluble solid, in were separated into the aqueous fraction, which DWCNTs were enriched. Reproduced with permission.[167] Copyright 2011, Royal Society while the less-functionalized DWCNTs and of Chemistry.

Adv. Mater. 2015, 27, 3105–3137



REVIEW

5.3. Reversible Covalent Chemistry

3117

www.advmat.de

REVIEW


3118

nanotubes through π–π stacking.[162] At low pH, protonated amines also interact with the defect sites of the nanotubes and at high pH, through amine adsorption.[162] It was observed that lysozyme selectively suspended large-diameter DWCNTs[32] and this is in agreement with computational studies where MWCNTs of larger diameter (40 nm) consistently showed stronger protein binding than those with a smaller diameter (10 nm).[164] This is also in agreement with studies of protein interactions with nanoparticles.[169] Nie et al. suggest that the stronger bundles Figure 6. Computer-generated molecular modeling of the interactions between nanotweezers formed by smaller-diameter nanotubes[32] and nanocalipers and appropriately sized SWCNTs. By tailoring the chiral diporphyrin structure may also contribute to their reduced solu- to a specific size and depth, the (6,5), (9,4) and (9,7) SWCNTs can be isolated. Reproduced [172] bilization via protein binding.[32] Because with permission. Copyright 2013, American Chemical Society. the large-diameter DWCNTs are selectively functionalized, centrifugation provides separation of the largernanotubes via centrifugation. Specificity arises from careful and smaller-diameter DWCNTs into supernatant and pellet, selection of the spacing between the calipers, which affects the respectively. TEM of the respective DWCNT samples reveals width of the host molecule and the depth at which the nanotwo distinct ranges of diameter, of 3–5 nm (Gaussian average of tube sits within the host.[171] These physical parameters become 4.0 nm) and 1–4 nm (Gaussian average of 2.7 nm). As diameter very important when targeting a specific nanotube, with larger is strongly correlated with wall number, DWCNTs can also be nanotubes requiring a deeper position within the host. separated from smaller-diameter SWCNTs, providing a convenFigure 6 shows an example of directly engineered diporphyient method for DWCNT enrichment of such samples. This was rins and their complimentary nanotube types. For example, the shown by adding a 4:1 mixture of pure DWCNTs and SWCNTs (6,5) nanotube (diameter of 0.76 nm) can be accommodated to a solution of lysozyme. The mixture was sonicated to aid by a nanotweezer structure. By altering the structure of the solubilization and then centrifuged, yielding sediment and a diporphyrin to have a slightly larger spacing, the (9,4) nanolysozyme-biofunctionalized supernatant. Only large-diameter tube (diameter of 0.92 nm) can be accommodated. In order to DWCNTs were found in the supernatant (as determined from accommodate nanotubes >1 nm, a nanocaliper structure can TEM), highlighting the applicability of this approach to selecbe used, which possesses arms that extend out to capture the tively produce fractions of large-diameter DWCNTs from such target nanotube. Such tailored host structures not only allow mixtures. for diameter sorting, but also for enantiomeric separation In the work of Dresselhaus and co-workers, long and according to the handedness of the specific nanotube.[171,172] By random single-stranded (ss-) DNA was employed to indiinclusion of a chiral center within the diporphyrin, two mirrorvidualize DWCNTs, forming suspensions stable at high pH image (S) and (R) hosts can be created for each target nanotube (>6.8–12.4).[160] As the pH becomes more acidic, the DNA diameter, resulting in separation of left-handed (LH) and righthanded (RH) nanotube enantiomers. Initially, the enthalpies begins to destabilize and then agglomerate due to protonaof association of the LH and RH nanotubes with an (S)-diportion of the backbone phosphate groups. It is hypothesized phyrin are approximately equal.[171] However, as the number that, as the larger-diameter nanotubes are more susceptible to bundling due to increased van der Waals forces, they agglomof diporphyrin molecules bound to the nanotube surface erate preferentially and can be separated through centrifugaincreases, the formation of the RH:(S) complex becomes more tion from the suspended small-diameter DWCNTs. However, energetically favorable than the LH:(S) complex. This means this is so far based solely on PL and optical absorption, which, that RH:(S) becomes more stable and therefore, more soluble, as previously discussed, provide less certainty in the case of than LH:(S). The opposite can be expected for the mirror-image DWCNTs without corroboration from the more-conclusive host, with LH:(R) becoming more stable. TEM analysis. DWCNT sorting via the use of chiral diporphyrins was reported by Liu et al., wherein DWCNTs were successfully sorted from MWCNTs and, to a lesser extent, SWCNTs.[33] This 5.5. Molecular Nanocalipers was made possible by tailoring a chiral diporphyrin into a nanocaliper structure with a spacer of 1.9 nm between porphyrins. The use of chiral diporphyrin nanotweezers and nanocalipers Extraction with the nanocaliper yielded significantly improved can achieve simultaneous discrimination of nanotube diameter, DWCNT purity (from 77% up to 90%) and narrower diammetallicity, and handedness in SWCNTs.[171,172] This technique eter distribution (from 1.23–3.23 nm to 1.25–2.75 nm). After removal of the nanocalipers via repeated washing of the centriinvolves tailoring of the diporphyrin structure to yield one that fuged solid with tetrahydrofuran (THF) and pyridine, circular is sterically compatible with a nanotube of specific diameter dichroism (CD) measurements were performed, where the and chiral angle. Such diporphyrins form stable host–guest presence of LH and RH DWCNTs species were confirmed. The complexes with the target nanotube through π–π and CH–π authors thus concluded that the DWCNT enantiomers are most interactions, which can be separated from the remaining



Adv. Mater. 2015, 27, 3105–3137


5.6. Density-Gradient Ultracentrifugation Density-gradient ultracentrifugation (DGU) is a technique originally designed for separating and isolating biological elements (such as macromolecules or viruses)[173] and was first used to separate SWCNTs by Hersam and co-workers in 2005.[58] This separation method involves taking individualized nanotubes suspended in surfactant and placing them into a graded fluid medium (usually iodixanol owing to its high viscosity and tunable density[142]) of varying density within a centrifuge tube.[58,130,140,143] The various nanotube species are then moved by centripetal forces until they reach their respective isopycnic points (the point where the nanotube's buoyant density matches that of the surrounding fluid medium). The key to the separation lies in the different buoyant densities of each nanotube species, which is a feature that is heavily reliant upon the physical structure of the nanotube itself, as well as the encapsulating surfactant and solvent molecules. In the case of a binary surfactant, the structure of the surfactant shell is strongly influenced by the nanotube diameter.[140,157] Alternatively, in co-surfactant mixtures the polarizability of the SWCNT plays an important role and leads to non-equivalent wrapping of the two surfactants around different nanotubes.[158] This is especially true for surfactant mixtures with SDS, which is highly sensitive to electronic type[140,158] and results in the encapsulated nanotubes having a strong contrast in density. For example, the addition of SDS to SCwrapped SWCNTs significantly increases the density of nanotubes with diameters of 0.83 nm and 0.98/1.03 nm, and facilitates their separation from smaller nanotubes (ca. 0.76 nm).[140] Therefore, by employing the appropriate conditions, SWCNT fractions with a narrow diameter distribution, or of defined

semiconducting and metallic type, or even of enantiomerically pure single chirality can be produced by DGU.[130,140,143] Due to the great success of DGU for SWCNTs, it was a natural step to extend the method to DWCNTs. In particular because the inclusion of an inner wall significantly alters the buoyant density and thus offers a convenient method to deal with unwanted SWCNT impurities. This was first achieved by Green and Hersam in 2009 using a two-step DGU process.[34] In the first step, raw material (DWCNT composition of 70%) is suspended in 1 wt% SC. The authors speculate that, due to the anionic planar structure of SC, its encapsulation layers are highly sensitive to nanotube diameter and therefore, its use affords fractions of SWCNTs, DWCNTs, and MWCNTs of very different densities.[34] Furthermore it is speculated that another advantage of SC is its limited sensitivity to the electronic character of the nanotubes, which means that any electronic perturbations of the outer wall by the inner wall will not have an impact on the surrounding surfactant layer. Therefore, for similar diameters, SC encapsulation does not vary between SWCNTs and DWCNTs, and the only difference is a much larger density for DWCNTs due to the inner wall. Figure 7a shows a schematic of the SC encapsulation of small- and largediameter SWCNTs and DWCNTs, as well as its effect on their buoyant densities. The density gradient consisted of 1.5 mL of 60 wt% iodixanol on the bottom, followed by 5 mL of a linear gradient varying from 32.5 to 17.5 wt% iodixanol with 1.5 mL of 15 wt% iodixanol on top, in which the DWCNTs initially resided. The remaining volume was filled with 0 wt% iodixanol and the linear density was then centrifuged at 41 000 rpm (207 570g) for 12 h. In this process the nanotubes are forced to sediment from their initial starting position of low density to a higher density. The advantage of top addition is that lower-density SWCNTs are prevented from reaching the isopycnic point of the much denser DWCNTs. As seen in Figure 7b, this results in 4 bands corresponding to small-diameter SWCNTs (ca. 0.7–1.1 nm),

REVIEW

probably obtained through molecular recognition by the chiral diporphyrin nanocalipers.

Figure 7. Separation of DWCNTs by the number of walls via DGU. a) Schematic illustration of nanotube encapsulation by SC and its effect on the resultant buoyant density. b) Photograph of a centrifuge tube following the first iteration of the separation showing four bands, corresponding to smalldiameter SWCNTs, large-diameter SWCNTs, DWCNTs, and MWCNTs/carbonaceous impurities/bundles. c) Absorption spectra of each band following the initial separation and, in the case of large-diameter SWCNTs and DWCNTs, the first iteration. The absorbance is normalized and offset in each case. Reproduced with permission.[34] Copyright 2009, Nature Publishing Group.

Adv. Mater. 2015, 27, 3105–3137



3119

www.advmat.de

REVIEW


Figure 8. Analysis of the sorting efficiency of the DGU technique by TEM. a) Centrifuge tube of the DGU sorted DWCNT material showing several bands within the linear gradient. b) Absorption spectra of each layer. c) TEM analysis of the different bands showing an increase in mean diameter with increasing density, change in relative concentration of SWCNTs and DWCNTs and the change in distribution shape. d–g) TEM images of SWCNTs and DWCNTs from layers 3, 6, 10 and 12, respectively. Reproduced with permission.[174] Copyright 2008, Wiley-VCH.

large-diameter SWCNTs (ca. 1.6 nm), DWCNTs (ca. 1.6 nm), and MWCNTs, bundles, and carbonaceous impurities. The coarsely refined DWCNTs and large-diameter SWCNTs were then subjected to a second DGU step. In that step the material was now inserted into the bottom of the density gradient, which forced the nanotubes to move from high to low density with the goal of removing any slow-moving, dense species that did not reach their isopycnic points the first step. The second density gradient consisted of 1.5 mL of 60 wt% iodixanol on the bottom, 1 mL of coarsely refined DWCNTs in 33.5 wt% iodixanol, then 5 mL of density gradient ranging from 31 to 16 wt% iodixanol with the remaining space filled with 0 wt% iodixanol. The effect of a subsequent DGU step is clearly visible in the absorption spectra in Figure 7c (dashed lines), where a significant improvement in the purity of the large-diameter SWCNT and DWCNT fractions is seen, as evidenced by the sharper peaks and reduced scattering background. In an investigation of the sorting efficiency of DGU, Loiseau and co-workers conducted a layer-by-layer analysis of nanotube content by TEM.[174] As seen in Figure 8a, twelve bands were taken from the centrifuge tube and characterized by TEM (over 100 nanotubes per band) and absorption spectroscopy (Figure 8b,c). As discussed,[34] the SWCNTs are found in the lower-density top half of the centrifuge tube with a trend of increasing diameter seen with increasing density (layers 1–5), whereas the highest-purity DWCNTs (85%) are found in layer 12. Between layers 6 and 12, a mixture of SWCNTs and DWCNTs were found with increasing DWCNT purity.

3120


Extending the DGU method to sort DWCNTs according to electronic character was a seminal advancement toward the realization of high-technology DWCNT devices, and was achieved by Green and Hersam in 2011.[35] In that work, DWCNTs coarsely purified via their original method[34] were separated by outer wall electronic type using subsequent DGU iterations. However, in the subsequent DGU steps, co-surfactant mixtures of SC and SDS were used. Due to the very different surfactant wrapping of metallic and semiconducting nanotubes,[158] the difference in density between the two electronic types in co-surfactant mixtures is even greater than in the single surfactant case and this can be optimized for a particular metallic/ semiconducting separation.[130,140,142] For instance, ratios of 1:4 SDS/SC and 3:2 SDS/SC have been found to be optimal for targeting large-diameter semiconducting and metallic SWCNTs, respectively.[140] For semiconducting DWCNT enrichment, three iterations followed the initial coarse DGU step. In the first iteration, the DWCNTs were placed at the bottom of a 1 wt% 1:4 SDS/ SC density gradient from 25 to 40 wt% iodixanol. The gradient was centrifuged (41 000 rpm or ca. 208 000g for 14 h) and resulted in further separation of DWCNTs and SWCNTs. However, separation according to electronic character was also seen owing to the introduction of SDS. The semiconducting outer wall enriched (s-) DWCNT layer was then isolated and used in the second iteration, which was a repetition of the first, but with the s-DWCNTs added to the top of the gradient. In the third iteration, a 3:2 SDS/SC ratio was used with the DWCNTs


Adv. Mater. 2015, 27, 3105–3137


REVIEW Figure 9. Separation of DWCNTs by outer wall electronic type. a–b) Photographs of centrifuge tubes following DGU separation of SWCNTs/DWCNTs targeted for semiconducting and metallic DWCNTs, respectively. c–d) Absorbance spectra of each band during the first-iteration separations targeting semiconducting and metallic DWCNTs, respectively. Metallic DWCNTs, semiconducting DWCNTs, metallic SWCNTs, semiconducting SWCNTs, and then coarsely enriched input DWCNT material are shown by the red, green, purple, blue, and dashed grey curves, respectively. The spectra are offset for clarity. e–f) Absorbance spectra obtained during successive DGU iterations to produce semiconducting and metallic DWCNTs, respectively. The asterisks mark the absorption peaks attributed to inner-DWCNT-wall transitions. Sii (Mii) label the ith order semiconducting (metallic) optical transitions of the SWCNTs and the outer walls of DWCNTs. Wavelength regions associated with semiconducting and metallic SWCNTs and outer wall DWCNTs transitions are shaded red and green, respectively. Reproduced with permission.[35] Copyright 2011, American Chemical Society.

moving from low to high density to remove any remaining metallic species. For metallic DWCNT enrichment, only two iterations followed the initial coarse DGU step and both used a 1 wt% 3:2 SDS/SC ratio, but differed in initial DWCNT placement. Figure 9a shows the resultant layer structure for a co-surfactant ratio of 1:4 SDS/SC. Separation occurs primarily by diameter, however red and green fringes corresponding to metallic and semiconducting character, can also been seen. The absorption measurements of each fringe (seen in Figure 9c) also show signs of electronic sorting for both SWCNTs and DWCNTs, with enhanced S22 and M11 transitions. Alternatively, the separation with a co-surfactant mixture of 3:2 SDS/SC can be seen in Figure 9b, where the increased relative concentration of SDS has significantly changed the resultant band structure. Separation is now dominated by electronic character and occurs concomitantly with diameter, with the four bands visible

Adv. Mater. 2015, 27, 3105–3137

corresponding to m-SWCNTs, m-DWCNTs, s-SWCNTs, and s-DWCNTs (absorption measurements shown in Figure 9d). As the m-CNTs are coated with an increased surface concentration of SDS compared to that of s-CNTs, their isopycnic points are at lower densities.[175,176] Figure 9e shows absorption spectra after the initial DGU step and the three subsequent iterations for s-DWCNT enrichment. While each DGU step yields enhancement of the S22 and reduction in M11 features, the most significant enrichment is seen in the final step, which employs the 3:2 SDS/SC co-surfactant mixture. Additionally, there is also a peak at ca. 1200 nm corresponding to the S11 transition of an inner wall. Figure 9f shows absorption spectra for each DGU step, targeting m-DWCNTs. Each separation sees the enhancement of the M11 transitions with S22 contributions from s-DWCNTs removed, revealing inner wall S11 peaks in the same region. In each case, very high purities were achieved with final fractions containing 96% and 98% semiconducting and metallic



3121

www.advmat.de

REVIEW


slightly lower density (30%, 1.16 g mL−1) containing the DWCNTs. Finally, 20 mL of 26% (1.14 g mL−1) solution was added (the race layer in which fractionation occurs), with the concentration of surfactant remaining constant throughout, at 1 wt% deoxycholate (DOC). Importantly, the density of the surrounding medium is greater than the average density of the DWCNTs (ca. 1.11 g mL−1), thus enabling transient motion and not buoyant density. The gradient was then centrifuged at 34 000g (16 640 rpm) for 49 h. The top-most layer (containing nanotubes that had travelled the furthest from the DWCNT Figure 10. Schematic and photographs of the process of length separation by centrifugation injection layer), contained nanotubes with a for SWCNTs. An injection layer containing the SWCNTs, modified to the appropriate liquid length of 2.2 µm. As expected, the average density, is placed near the bottom of the tube to maximize the separation. Longer nanotubes length decreased for fractions closer to the move further in response to the applied centrifugation, and thus separate up the tube; any high- injection layer, with the shortest average density impurities move in the opposite direction. Reproduced with permission.[141] Copyright length reported as 0.6 µm, with each layer 2008, American Chemical Society. having a relatively narrow length distribution (distribution of layer 4 was ±0.18–0.3 µm). Fagan and co-workers also report partial enrichment by elecDWCNTs, respectively. This clearly demonstrates the effectivetronic character using co-surfactants.[178] In that case, a slightly ness of the DGU technique. denser gradient was employed with an additional co-surfactant gradient; however, evidence of electronic enrichment was limited. 5.7. Centrifugal Length Separation In addition to the separation of carbon nanotubes by diameter and electronic type, centrifugation has also been used to prepare nanotubes with a narrow length distribution.[141,177] This length-dependent separation technique has previously been reported for the fractionation of SWCNTs with lengths in excess of 1 µm and is achieved by exploitation of the lengthdependent friction coefficient through a dense liquid under centrifugation.[141,177] While this technique is very similar to DGU, it employs a fluid medium much denser than the nanotubes such that they can never reach their isopycnic points, and, rather, exploits transient motion, where longer nanotubes travel with greater velocity in opposition to the applied acceleration. This occurs because the rate of nanotube flow through the fluid has a non-linear dependence on length.[177] Separation occurs (with minimal chirality differentiation), provided that Δρ = ρS – ρSWCNT >> ΔρSWCNT = ρSWCNT – ρSWCNT,i where Δρ is the difference in density, ρS is the density of solution, ρSWCNT is the average density of all the SWCNT chiralities and ρSWCNT,i is the density of an individual SWCNT chirality.[177] Thus, the key to length separation is to choose the linear density such that Δρ >> ΔρSWCNT, thus exploiting the transient motion regime and not the regime in which buoyancy equilibrium is reached. Figure 10 shows a schematic representation and photographs of SWCNT length separation by centrifugation, where longer SWCNTs travel further from the injection layer. In 2010, Fagan and co-workers separated DWCNTs according to length in a two-step method.[178] Firstly, the DWCNTs were coarsely enriched using the pre-established DGU method.[34] The enriched DWCNT material was then inserted into a centrifuge tube with a specially designed gradient derived from SWCNT length separations.[141,177] Then, 1 mL of high-density (40%, 1.21 g mL−1) iodixanol was added, followed by 1 mL of

3122


5.8. Gel Permeation The use of Sephacryl gel permeation, first demonstrated by Moshammer et al.,[59] has been shown to be extremely successful in the preparation of (n,m)-purified SWCNT suspensions. For SWCNTs, this technique has allowed for the high-throughput separation of metallic species from semiconducting species,[59,152,153] the isolation of specific (n,m) species,[60–62] and, most recently, the separation of optical isomers.[179] In general, the gel-permeation method involves passing SDS-suspended nanotubes through a stationary phase gel bed contained within a column, at which point semiconducting nanotubes with the highest affinity for the gel (an interaction determined by the SDS wrapping, and hence, the individual nanotube structure) are selectively removed from the bulk solution and retained on the gel matrix.[61] The metallic nanotubes, which exhibit no interaction with the gel,[60,61,152,153] as well as other semiconducting nanotubes with no affinity to the gel (determined by surfactant concentration), continue to flow through the gel and can be collected. The gel is then washed with SDS of either increased concentration[60,61,152] or lower pH,[153,180] or alternatively SC,[59,152] which disrupts the interaction between the adsorbed semiconducting nanotubes and the gel and thus elutes them from the column for collection. This process can be repeated sequentially with the nanotubes that have the highest affinity for the gel becoming preferentially adsorbed each time. While the exact mechanism of gel-based separation remains speculative owing to the difficulties associated with determining the molecular dynamics on the nanoscale, evidence suggests that it is a kinetically driven selective adsorption


Adv. Mater. 2015, 27, 3105–3137


Extensive atomic force microscopy (AFM), absorption spectroscopy, TEM, and Raman spectroscopy revealed that those bands corresponded to DWCNTs and SWCNT, respectively. One explanation for the significant difference in elution rates is that the nanotubes undergo a size-exclusion process in which the retention time is dependent on the size. Indeed, extensive AFM revealed that there was a significant size difference between the two nanotubes types, which is summarized in Figure 11b. On average, the DWCNTs (ca. 700 nm) are twice as long as the SWCNTs (ca. 300 nm), which may be attributed to the increased structural stability of the DWCNTs during extensive sonication. This is in agreement with Green and Hersam who also observed that DWCNT lengths were ca. 40% longer than SWCNTs after 1.5 h of sonication.[34] However, a true sizeexclusion process would also occur in SC, which the authors confirm is not the case. The mechanism is therefore likely a combination of two competing factors: diameter-dependent solvation by the surfactant and length-dependent size exclusion. Analogous to the evolution of DWCNT sorting by the DGU method, the gel-permeation technique was then extended to the electronic sorting of DWCNTs.[37] In that work, a DWCNT suspension in 1 wt% SC was added to a 20 cm long gel column in 1 wt% SDS. Upon addition of the SC-encapsulated nanotubes to the column, they immediately underwent surfactant exchange, resulting in various ratios of SDS/SC wrapping, depending on the diameter and electronic character.[130,142,143] Owing to this difference in wrapping, nanotubes of different electronic types moved through the column at different rates, forming four bands, which were eluted from the column after different retention times. The four bands were shown to correspond to defected m-DWCNTs, m-DWCNTs, s-DWCNTs, and s-SWCNTs, respectively, with the elution order in excellent agreement with data from the separation of AD and HiPco SWCNTs of a similar electronic type. Thus, it is clear that the separation process is highly sensitive to the surface properties of each nanotube type, with the presence of the inner wall having little effect on the separation. The authors suggest that the mechanism is similar to that observed previously for SWCNT separation, where SDS is sensitive to electronic character and diameter.[60,61,152] Metallic nanotubes, which are known to have a stronger interaction with SDS compared to semiconducting nanotubes,[175] become more fully wrapped and experience limited interaction with the gel as they traverse the column. Thus, they elute first, followed by largeand small-diameter semiconducting nanotubes. Figure 12a shows the elution profile of DWCNTs, large-diameter AD SWCNTs and small-diameter HiPco SWCNTs. An excellent agreement between elution times is observed for metallic (Band 2) and semiconducting (Band 3) fractions of DWCNTs and AD SWCNTs, as well as between HiPco SWCNTs and SWCNTs within the DW sample. Absorption spectra of the enriched Figure 11. Separation of DWCNTs by gel permeation. a) Time-lapse photography (1 h) showing m- and s- DWCNT fractions are shown in the introduction of raw, unsorted DWCNT material to the S-200 gel column for the separation of SWCNTs from DWCNTs and b) diameter vs length of SWCNTs and DWCNTs, as determined Figure 12b,c, demonstrating clear enhancement of the M11 (600–800 nm) and S22 by AFM. Reproduced with permission.[36] Copyright 2013, American Chemical Society.

Adv. Mater. 2015, 27, 3105–3137



REVIEW

process,[61] highly dependent on the SDS wrapping of the nanotubes.[151] This is evidenced by the clear relationship between SDS concentration and gel adsorptivity,[60,61,151,152] where an increase in the SDS concentration allows additional SDS molecules onto the nanotube surface,[176] reducing its interaction with the gel. In addition to the SDS concentration, pH[62,153] and temperature[151] have also been shown to play important roles, giving rise to a number of separation strategies. While SDS-based separation is successful for small-diameter SWCNTs (0.77–1 nm), adsorption of large-diameter nanotubes (greater than ca. 1.2 nm) to the gel is quite limited.[37] Instead, a co-surfactant separation method can be used, where either the largediameter material is suspended in a co-surfactant solution and applied to a column, as demonstrated by Miyata et al.[181] and Wu et al.,[182] or where they are suspended in SC and applied to a column in SDS, as demonstrated by Zhang et al.[183] While all of these co-surfactant methods produced highly pure, semiconducting solutions (99%, 98%, and 98%, respectively), none report the purity of the metallic fraction, which is washed off of the gel in the flow-through. Application of the gel-permeation technique to DWCNTs was first reported by Moore et al. in 2014,[36] wherein the separation of DWCNTs from SWCNTs was demonstrated, producing highly enriched fractions with mean diameters of 1.64 ± 0.15 and 0.93 ± 0.03 nm, respectively. In that work, DWCNT raw material was suspended in 2 wt% SDS by sonication for 8 h, and this was then added to a 25 cm Sephacryl S-200 gel column, also at 2 wt% SDS. While the bulk of the nanotubes passed through the gel, a small amount remained adsorbed to the top of the column. This is in agreement with previous work by Blanch et al.[152] and Flavel et al.[62] where it has been demonstrated that, at a relatively high SDS concentration (1.6–2 wt%), only a very small portion of the overall nanotube population is adsorbed in a competitive process, compared to that for low SDS concentrations (0.4–0.8 wt%). Addition of 0.5 wt% SC resulted in the desorption of the nanotube material and two distinct bands formed in the column, the first of which was fast moving and dispersed while the second was slower moving and more tightly confined. This can be seen in Figure 11a, which shows time-lapse photographs of the initial adsorption and subsequent elution with SC.

3123

www.advmat.de

REVIEW


Figure 12. The co-surfactant separation of DWCNTs via gel permeation. a) Elution profiles of the normalized G-Band Raman mode intensity for DWCNTs, AD SWCNTs and HiPco SWCNTs. The dashed lines in the DWCNT elution profile highlight Band 2 and Band 3, which, from the absorption spectra seen in (b) and (c), correspond to DWCNTs with metallic and semiconducting outer walls. Regions of Sii and Mii transitions are highlighted in each spectrum. Reproduced with permission.[37] Copyright 2015, American Chemical Society.

(900–1200 nm) transitions with calculated purities of ca. 70% and ca. 90%, respectively.

5.9. Polymer Wrapping The dispersion and separation of SWCNTs with aromatic polymers in organic solvents is receiving ever-increasing attention in the research community due to the ability of polymer wrapping to prepare suspensions with a highly pure semiconducting content (>99%).[184–186] In this one-pot approach, raw carbon nanotube material is typically dispersed by ultrasonication in the presence of a chosen polymer followed by ultracentrifugation. Consequently, only those nanotubes with a preferential interaction with the polymer and those that have become individualized during ultrasonication remain suspended in the organic solvent. In pioneering works by Nish et al.,[147] Chen et al.,[148] and Hwang et al.,[132] organic polymers with the fluorene structure as part of their repeat unit were used, such as poly(9,9-dioctylfluorene-2,7-diyl) (PFO), poly[9,9-dihexylfluorenyl-2,7-diyl)-co-(9,10-anthracene)] (PFH-A) and poly[(9,9dioctylfluorenyl-2,7-diyl)- co -1,4-benzo-[2,1’-3]-thiadiazole)] (PFO-BT). After almost 10 years of development, the polymer library has grown dramatically to now include polythiophenes, polycarbazoles, and copolymers thereof,[39,187–192] alongside research to develop new polymers via click chemistry in an effort to avoid the strict synthetic conditions associated with Suzuki polycondensation or Yamamoto coupling.[193] Using the

3124


currently available polymer library, mixtures of semiconducting SWCNTs through to near-monochiral and monochiral suspensions and most recently even optical isomers thereof [194] have been demonstrated. For example, poly(9,9-dialkyl-2,7-fluorene) has been shown to be sensitive to large chiral angles (close to armchair: θ ≥ 25°)[132,148,195] and poly(N-decyl-2,7-carbazole) to lower chiral angles (typically 10° ≤θ ≤ 20°).[191] Despite the wide-spread application of polymer wrapping, the exact mechanism responsible for separation remains poorly understood and is the subject of current discussion.[185,186,193,196] Nevertheless, it is commonly agreed upon that the polymer interacts by aligning its aromatic backbone along the surface of a carbon nanotube so as to maximize π–π stacking.[38,39,185,188,190] Peripheral groups are then believed to branch away from the nanotube into the solvent and facilitate solubility.[186,196] Berton et al.[187] have developed a hybrid coarse-grain model to describe this interaction by treating the nanotube and polymer as geometrical objects and assuming maximal (attractive) π–π interactions can be reached by maximizing the contact area. In this way, a set of solutions for which the polymer lies flat on the surface of the nanotube were calculated as shown in Figure 13 for poly(fluorene-alt-pyridine) around a 1.2 nm-diameter SWCNT. Intuitively, and also in agreement with Berton's model, the number of favorable wrapping solutions increase with diameter, which may explain why certain polymers are highly (n,m) specific in the small-diameter regime (i.e., due to other species having limited possible wrapping solutions).


Adv. Mater. 2015, 27, 3105–3137


Adv. Mater. 2015, 27, 3105–3137



REVIEW

In the past, research efforts focused mainly on the dispersion of small-diameter (0.8–1.2 nm) carbon nanotubes from the CoMoCat or HiPco processes, and selectivity was shown to be poor for larger diameters.[38,187,197] As the research community now turns to larger-diameter nanotubes due to their potential application in telecommunications[198] and FETs,[199] new polymer systems are also beginning to appear in the literature. Initially it was believed that the poor selectivity to large diameters could be ascribed to Figure 13. One of the geometric solutions for poly(fluorene-alt-pyridine) wrapped around a the nature of the polymer backbone; however, 1.2 nm-diameter SWCNT and the corresponding molecular model. Alkyl chains are simplified recent studies point to side chains as playing as methyl groups. The red discs and black dots represent the fluorene and pyridine moieties, an important role.[38,192,197,200] Gomulya and and these two objects are connected by sticks whose lengths reflect the bonds between the co-workers[38,39] investigated the effect of fluorene moiety and the center of the pyridine ring. Adapted with permission.[187] Copyright using polyfluorene derivatives with different2014, Wiley-VCH. length alkyl side chains from C6H13 up to C18H37, and found that an alkyl chain length of 8 carbons favored dispersion of diameters of 0.8–1.2 nm, The observed high selectivity of polymer systems to semiwhereas longer alkyls with 12–15 carbons can efficiently conducting SWCNTs is much less understood and experiinteract with nanotubes with diameters up to 1.5 nm. Unformental evidence suggests that it is dependent upon a number tunately, in spite of improved sensitivity to large diameters, the of factors, such as polymer concentration,[186] ultrasonication longer alkyl chains also had the effect of increasing the intertemperature,[185] and solvent choice.[132,147,185,196] Changes in action strength with the nanotubes, and thereby decreased the any one of these parameters have been shown to influence not selectivity. In other work, Berton et al.[188] reported the disperonly the concentration of metallic SWCNTs in solution but also the polydispersity of the semiconducting species. For example sion of 1.3 nm-diameter SWCNTs using poly(9,9-didodecylfluShea et al.[186] showed that for high starting concentrations of orene-2,7-diyl-alt-anthracene-1,5-diyl); Tange et al.[201] dispersed PFO in toluene (5 mg mL−1) the polydispersity of s-SWCNT 1.3 to 1.4 nm diameters with poly(9,9-dioctylfluorene-alt-benzothiadiazole) (F8BT), and Wang et al.[197] dispersed AD carbon (n,m) species increased from 5 to 8 and m-SWCNTs became visible in optical absorption measurements. Conversely, at nanotubes with poly(dithiafulvalene-fluorene-co-m-thiophene). polymer concentrations below 1 mg mL−1, no m-SWCNTs were To date, polymer wrapping has yet to be applied to DWCNTs, but, considering that the outer wall of a DWCNT has a diamsuspended. As stated by the authors, the increased polydispereter of greater than ca. 1.5 nm, these newly developed polymers sity and metallic content is likely to be a result of high PFO conlook highly promising for the future of DWCNT processing. centrations suspending not only individualized SWCNTs but However, as several researchers point out, the polymer wrapalso polydisperse bundles. Likewise, Han et al.[185] dispersed ping of a carbon nanotube is also associated with a charge carbon nanotubes with PFO in cyclohexane at increasing temtransfer,[190,196,197] and what effect the presence of an electroniperature (5–55 °C) and found that increased ultrasonication temperature led to an overall higher concentration of s- and cally coupled inner wall may have on the sorting efficiency m- SWCNTs. remains to be seen. However, it is the choice of solvent that has captured the attention of many in the research community,[132,147,185,196] as it has dramatic effects on the semiconducting purity. As out5.10. Aqueous Two-Phase Extraction lined by Wang et al.,[196] upon selecting a solvent, several rules Another separation method that looks promising for the separaare typically adhered to; namely, the solvent must solubilize the tion of DWCNTs is an adaptation of standard liquid-phase sepapolymer, the SWCNTs must have a low intrinsic solubility in the ration, recently applied to the separation of carbon nanotubes solvent so that only polymer-wrapped nanotubes are dispersed, by Zheng and co-workers.[40] In this first report, SC-dispersed and the solvent must have a lower density than the SWCNTs so that the unwrapped SWCNT will sediment after centrifugananotubes (and an appropriate amount of SDS) were added to tion. Additionally, Wang et al.[196] go further to state that nona mixture of two water-soluble polymers; polyethylene glycol (PEG) and dextran. Spontaneous and robust separation of the polar solvents are necessary to prevent solvent interactions with polymer phases occurred with the more-hydrophobic PEG-rich polarized polymer-wrapped m-SWCNTs and allow for better phase on top and the more-hydrophilic dextran-rich phase on selective sorting of s-SWCNTs. As an example, the use of conthe bottom. The nanotubes quickly and spontaneously sepajugated polymers in polar solvents such as THF, typically have rated into the two polymer phases with thermodynamic anala higher dispersive yield (i.e., a higher final suspended mass of ysis revealing two distinct regimes. First, in the small-diameter CNTs) but suspend both m- and s- SWCNT, whereas nonpolar regime (0.6–1.0 nm), curvature dominates the solvation free solvents such as toluene, o-xylene, and m-xylene, whilst having energy with smaller-diameter nanotubes such as (6,4) in the a lower dispersive yield, have been shown to afford semiconmore-hydrophilic dextran-rich phase, and larger-diameter ducting purities of >99%.[184–186]

3125

www.advmat.de

REVIEW

www.MaterialsViews.com Table 1. Comparison of the advantages and disadvantages of all current DWCNT-sorting methods with a focus toward commercial application. Technique Reversible covalent chemistry

Advantages

Disadvantages

Diameter selective

Cannot distinguish between number of walls

Easily scalable

Currently insensitive to electronic character

Pristine structure can be recovered

Multiple steps; time consuming

Scalable Biofunctionalization

Molecular nanocalipers

Diameter selective

Cannot distinguish between SWCNTs/MWCNTs and DWCNTs

Easily scalable

Not sensitive to electronic character

Diameter selective

Requires complex chemical engineering to make hosts

Sensitive to handedness

Requires expensive reagents

Pristine nanotubes can be recovered

Limited evidence of electronic sensitivity

Host can be reused DGU

Diameter selective

Requires expensive ultracentrifuge

Sensitive to electronic character

Requires expensive density gradient medium

Can distinguish between number of walls

Requires technical expertise to make intricate density gradients Multiple steps; time consuming

Centrifugal separation

Length selective

Requires pre-sorted DWCNT material Limited evidence of electronic sensitivity Requires expensive density gradient medium

Gel Permeation

Diameter selective

Requires expensive gel

Sensitive to electronic character Polymer wrapping

Diameter selective

Requires centrifuge

Sensitive to electronic character

Requires expensive polymers

Scalable Aqueous two-phase extraction

Diameter selective

Diameter selectivity currently limited to small diameter regime < 1.2 nm

Sensitive to electronic character Easily scalable Inexpensive, readily available reagents True “one-pot” method

nanotubes such as (7,5) and (8,4) in the more-hydrophobic PEGrich phase. The second, larger-diameter regime (1.2–1.5 nm), is governed by the degree of nanotube polarizability, and a clean metallic/semiconductor separation can be achieved, with metallic nanotubes found in the more-hydrophilic dextran-rich phase and semiconducting nanotubes in the more-hydrophobic PEG-rich phase.[40] Investigation by Subbaiyan et al. determined that aqueous two-phase separation was driven by the hydrophobicity of the surfactant composition on the nanotube surface rather than the inherent hydrophobicity of the nanotube itself,[41] therefore allowing the surfactant identity and ratio to be tuned to target specific chiralities in a similar manner to gel permeation and DGU. By doing so, it is possible to enrich (6,5), (6,4)/(7,3) and (7,5) species. Since then, several advances have been made through variation of the surfactant concentration[202] and the type of polymer,[203] addition of salts[202] or DNA,[203] and the use of countercurrent chromatography.[204] Single-chirality nanotube suspensions can now be prepared for a number of species, and the use of countercurrent chromatography enables total fractionation with a recovery close to

3126


90% of the starting material.[204] While this rapidly developing technique has made extraordinary advances for small-diameter (

Carbon nanotube biosensors.

Processing and Characterization of a Novel Distributed Strain Sensor Using Carbon Nanotube-Based Nonwoven Composites.

Structural Polymer-Based Carbon Nanotube Composite Fibers: Understanding the Processing-Structure-Performance Relationship.

Carbon Nanotube-Based Chemiresistive Sensors.

An Organocobalt-Carbon Nanotube Chemiresistive Carbon Monoxide Detector.

Carbon Nanotube Flexible and Stretchable Electronics.

How a zigzag carbon nanotube grows.

Aqueous nanosilica dispersants for carbon nanotube.

A Thermal Model for Carbon Nanotube Interconnects.

Fermentation based carbon nanotube multifunctional bionic composites.

Progress Towards Applications of Carbon Nanotube Photoluminescence.

Copper-encapsulated vertically aligned carbon nanotube arrays.

Carbon nanotube alignment driven rapid actuations.

Molecular ionization from carbon nanotube paper.

Printed Carbon Nanotube Electronics and Sensor Systems.

Free-standing carbon nanotube-titania photoactive sheets.

Fibrillous carbon nanotube: an unexpected journey.

Polychiral semiconducting carbon nanotube-fullerene solar cells.

Synthesis of Carbon Nanotube (CNT) Composite Membranes.

Applications of three-dimensional carbon nanotube networks.

polymer composites.

Fitting Single-Walled Carbon Nanotube Optical Spectra.

Antimicrobial Activity of Chitosan-Carbon Nanotube Hydrogels.

Bio-inspired Hybrid Carbon Nanotube Muscles.